YOLOv8目标检测在RK3588部署全过程

于 2025-02-17 14:52:15 发布
阅读量528 收藏 3
点赞数
分类专栏： 机器学习 深度学习 文章标签： YOLO 目标检测 人工智能
原文链接：https://blog.youkuaiyun.com/qq_44764044/article/details/140062461
版权
说明：该文章为引用，方便自己后续查看，如有需要请移步原作者网址：（https://blog.youkuaiyun.com/qq_44764044/article/details/140062461），如有侵权，请联系删除。谢谢。

一，前言

这是一个关于从电脑安装深度学习环境到实现YOLOv8目标检测在RK3588上部署的全过程。

本人配置：

1，一台笔记本

2，一个香橙派5s

二，深度学习环境配置

2.1 安装anaconda

使用清华镜像源下载https://mirrors.tuna.tsinghua.edu.cn/anaconda/archive/根据自己的电脑选择下载安装

安装之后把下面几个Anaconda的环境变量添加进系统变量中。

2.2 安装Pycharm

去Pycharm官网下载https://www.jetbrains.com/pycharm/download/，有专业版（收费）和社区版。

2.3 安装CUDA与cuDNN

参考https://blog.youkuaiyun.com/zyq880625/article/details/138086225

2.3.1安装CUDA

win+R打开cmd，输入nvidia-smi来查看自己可以安装的CUDA版本（不能超过12.2）。

nvidia-smi

  
  

  
  
  • 1

CUDA下载地址CUDA Toolkit Archive | NVIDIA Developer，选择一个不高与自己CUDA的版本，然后进入选择自己电脑的配置最后下载安装。

安装过程
1、双击下载的EXE安装包，开始安装；
2、提取安装文件的（临时）存放位置，保持默认，点击OK，等待文件提取完成；
3、等待检查系统兼容性；
4、许可协议，点击同意并继续；
5、如果是第一次安装，选择精简（精简版本是下载好所有组件，并且会覆盖原有驱动），一直点击下一步即可，安装完成关闭即可；

查看是否安装成功

在cmd中输入nvcc -V

nvcc -V

  
  

  
  
  • 1

外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传

2.3.2 安装cuDNN

去cuDNN Archive | NVIDIA Developer选择与自己CUDA合适的版本下载，下载后将压缩包解压

将解压后的文件夹中的bin、include、lib 三个文件夹，移动到CUDA Development 安装路径下，
与同名文件夹合并。
CUDA Development 默认安装路径：

C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0

  
  

  
  
  • 1

添加至系统变量：

往系统环境变量中的 path 添加如下路径（根据自己的路径进行修改）

C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\bin
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\include
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\lib
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\libnvvp

  
  

  
  
  • 1
  • 2
  • 3
  • 4

2.4 安装GPU版本的Pytorch

打开 Anaconda prompt

使用 conda create -n 虚拟环境名字 虚拟环境的python版本 来创建虚拟环境

conda create  -n pytorch python=3.9

  
  

  
  
  • 1

创建完虚拟环境后，再激活虚拟环境,，左边括号表示现在所处的环境

conda activate pytorch

  
  

  
  
  • 1

外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传

pytorch 的下载链接：download.pytorch.org/whl/torch_stable.html，torch，torchvision的对应关系如下

因为我创建的pytorch环境的python是3.9所以我选择的是torch=2.0.0，torchvision=0.15.1

下载完成后在Anaconda prompt使用下面命令安装

pip install torch-2.0.0+cu118-cp39-cp39-win_amd64.whl
pip install torchvision-0.15.1+cu118-cp39-cp39-win_amd64.whl

  
  

  
  
  • 1
  • 2

安装完成后在Anaconda prompt中打开python，使用下面命令查看是否可以使用GPU。

import torch
torch.cuda.is_available()

  
  

  
  
  • 1
  • 2

返回True则表示安装成功。

外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传

三，训练自己的模型

3.1 YOLO环境配置

下载瑞星微提供的工程https://github.com/airockchip/ultralytics_yolov8.git，不要使用YOLO官方提供的，下载完成后解压

使用Pycharm打开刚才解压的工程文件如下所示，然后点击右下角选择添加解释器

找到自己Anaconda的位置，然后选择自己创建的环境的解释器。

外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传

在下方点开终端选择Command Prompt就可以执行Anaconda prompt下的操作

外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传

使用下面的命令安装YOLOv8需要的依赖

pip install ultralytics 

  
  

  
  
  • 1

外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传

3.2 训练自己的数据集

3.2.1 数据集制作

在工程下面新建一个my_dataset文件夹用来存放数据集的图片images和标签labels。train和va和test的比例通常为8：1：1

my_dataset
│
└───images
    │   train
    │   val
    │   test
│   
└───labels
    │   train
    │   val

  
  

  
  
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10

在这里插入图片描述

若有其他格式的标签可以网上找代码转换成YOLO格式的，也可以使用labelimg进行数据集进行标注

安装labelimg

pip install labelimg

  
  

  
  
  • 1

在终端输入下面命令直接打开

labelimg

  
  

  
  
  • 1

外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传

在这里插入图片描述

标记完成后可以打开标签文件查看

外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传

3.2.2 修改配置文件

修改ultralytics/cfg/models/v8/yolov8.yaml文件中的nc为自己的类别数

ultralytics_yolov8-main/ultralytics/cfg/datasets下新建my_dataset.yaml，与文件夹下其他yaml内容文件类似指明训练，验证的路径和自己类别的标签

path: E:/multimodal_learning/Model_deployment/yolov8_test/ultralytics_yolov8-main/my_dataset # dataset root dir
train: ./images/train
val: ./images/val
test: ./images/test

# Classes 修改为自己的标签
names:
0: cat

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8

外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传

3.2.3 开始训练

如果想要使用YOLO官网的预训练权重的可以在Detect - Ultralytics YOLO Docs 下载预训练权重，此处下载的是YOLOv8n的权重文件

在根目录下新建一个train.py使用下面的程序，并把下载的预训练权重文件移动过来，鼠标右键后点击训练

from ultralytics import YOLO

if name == ‘main’:
model = YOLO(‘./ultralytics/cfg/models/v8/yolov8.yaml’).load(“./yolov8n.pt”)#创建模型并加载预训练权重

<span class="token comment"># Train the model</span>
model<span class="token punctuation">.</span>train<span class="token punctuation">(</span>epochs<span class="token operator">=</span><span class="token number">100</span><span class="token punctuation">,</span> batch<span class="token operator">=</span><span class="token number">8</span> <span class="token punctuation">,</span>data<span class="token operator">=</span><span class="token string">'./ultralytics/cfg/datasets/my_dataset.yaml'</span><span class="token punctuation">)</span>
<span class="token comment">#epochs表示训练的轮数，batch表示每次加载图片的数量，data是数据集的路径这里使用的是之前创建的yaml文件</span>
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8

训练结束R结果保存在 runs\detect\train下面，可以点开验证图片进行查看，weights保存着best.pt是我们训练出来的最好的权重文件
外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传

可以新建一个predict.py来测试，加载自己训练的权重文件对测试图片进行预测，保存路径会打印出来

from ultralytics import YOLO

model = YOLO(“./runs/detect/train/weights/best.pt”) # load a custom model

# Predict with the model
results = model(“./my_dataset/images/test/000000000650.jpg”,save=True)

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6

3.3 导出onnx模型

新建一个py文件运行后会在 runs\detect\train\weights下生成best.onnx文件，

from ultralytics import YOLO

# Load a model
model = YOLO(‘./runs/detect/train/weights/best.pt’)

# Export the model
model.export(format=‘rknn’,opset=12)

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8

可以使用Netron来查看自己的模型，这里查看下导出的onnx文件。

在这里插入图片描述

四，模型转换

4.1 转换环境安装

把onnx模型转换成rknn模型需要一个linux环境的电脑，如果没有可以在电脑上安装虚拟机，这里使用一台Ubuntu环境的服务器。

在服务器中也需要安装Anaconda和Pycharm与前面2.1节和2.2节类似，这里不再赘述。

先创建一个rk3588的虚拟环境

conda create  -n rk3588 python=3.9

 
 

 
 
  • 1

进入rk3588环境

conda activate rk3588

 
 

 
 
  • 1

下载rknn-toolkit2https://github.com/airockchip/rknn-toolkit2.git我选择的1.6.0的版本

外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传

解压后进入rknn-toolkit2-1.6.0/rknn-toolkit2/packages/中

外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传

安装依赖

pip install -r requirements_cp39-1.6.0.txt -i https://mirror.baidu.com/pypi/simple

 
 

 
 
  • 1

安装rknn_toolkit2

pip install rknn_toolkit2-1.6.0+81f21f4d-cp39-cp39-linux_x86_64.whl

 
 

 
 
  • 1

安装完成后进入python，使用下面命令没有报错，则安装成功。

from rknn.api import RKNN

 
 

 
 
  • 1

外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传

4.2 转换成rknn模型

新建一个yolov8_test的文件夹来存放我们的文件

best.onnx是我们3.3导出的onnx模型，dataset下面放一张测试图片000000000650.jpg和dataset.txt，txt中写的是测试图片的名字。

dataset-1下面放的是我们要测试的图片，测试的结果会保存在dataset-2下。

.
├── best.onnx
├── dataset
│   ├── 000000000650.jpg
│   └── dataset.txt
├── dataset-1
│   └── 000000000650.jpg
├── dataset-2
└── inference_test-2.py

 
 

 
 
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9

inference_test-2.py内容如下，包含模型转换和python部署的后处理

import os
import cv2
from rknn.api import RKNN
import numpy as np

IMG_FOLDER = “dataset-1”
RESULT_PATH = ‘./dataset-2’

#修改为自己的类别
CLASSES = [‘cat’]

OBJ_THRESH = 0.45
NMS_THRESH = 0.45

MODEL_SIZE = (640, 640)

color_palette = np.random.uniform(0, 255, size=(len(CLASSES), 3))

def sigmoid(x):
return 1 / (1 + np.exp(-x))

def letter_box(im, new_shape, pad_color=(0, 0, 0), info_need=False):
# Resize and pad image while meeting stride-multiple constraints
shape = im.shape[:2] # current shape [height, width]
if isinstance(new_shape, int):
new_shape = (new_shape, new_shape)

<span class="token comment"># Scale ratio</span>
r <span class="token operator">=</span> <span class="token builtin">min</span><span class="token punctuation">(</span>new_shape<span class="token punctuation">[</span><span class="token number">0</span><span class="token punctuation">]</span> <span class="token operator">/</span> shape<span class="token punctuation">[</span><span class="token number">0</span><span class="token punctuation">]</span><span class="token punctuation">,</span> new_shape<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">]</span> <span class="token operator">/</span> shape<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">]</span><span class="token punctuation">)</span>

<span class="token comment"># Compute padding</span>
ratio <span class="token operator">=</span> r  <span class="token comment"># width, height ratios</span>
new_unpad <span class="token operator">=</span> <span class="token builtin">int</span><span class="token punctuation">(</span><span class="token builtin">round</span><span class="token punctuation">(</span>shape<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">]</span> <span class="token operator">*</span> r<span class="token punctuation">)</span><span class="token punctuation">)</span><span class="token punctuation">,</span> <span class="token builtin">int</span><span class="token punctuation">(</span><span class="token builtin">round</span><span class="token punctuation">(</span>shape<span class="token punctuation">[</span><span class="token number">0</span><span class="token punctuation">]</span> <span class="token operator">*</span> r<span class="token punctuation">)</span><span class="token punctuation">)</span>
dw<span class="token punctuation">,</span> dh <span class="token operator">=</span> new_shape<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">]</span> <span class="token operator">-</span> new_unpad<span class="token punctuation">[</span><span class="token number">0</span><span class="token punctuation">]</span><span class="token punctuation">,</span> new_shape<span class="token punctuation">[</span><span class="token number">0</span><span class="token punctuation">]</span> <span class="token operator">-</span> new_unpad<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">]</span>  <span class="token comment"># wh padding</span>

dw <span class="token operator">/=</span> <span class="token number">2</span>  <span class="token comment"># divide padding into 2 sides</span>
dh <span class="token operator">/=</span> <span class="token number">2</span>

<span class="token keyword">if</span> shape<span class="token punctuation">[</span><span class="token punctuation">:</span><span class="token punctuation">:</span><span class="token operator">-</span><span class="token number">1</span><span class="token punctuation">]</span> <span class="token operator">!=</span> new_unpad<span class="token punctuation">:</span>  <span class="token comment"># resize</span>
    im <span class="token operator">=</span> cv2<span class="token punctuation">.</span>resize<span class="token punctuation">(</span>im<span class="token punctuation">,</span> new_unpad<span class="token punctuation">,</span> interpolation<span class="token operator">=</span>cv2<span class="token punctuation">.</span>INTER_LINEAR<span class="token punctuation">)</span>
top<span class="token punctuation">,</span> bottom <span class="token operator">=</span> <span class="token builtin">int</span><span class="token punctuation">(</span><span class="token builtin">round</span><span class="token punctuation">(</span>dh <span class="token operator">-</span> <span class="token number">0.1</span><span class="token punctuation">)</span><span class="token punctuation">)</span><span class="token punctuation">,</span> <span class="token builtin">int</span><span class="token punctuation">(</span><span class="token builtin">round</span><span class="token punctuation">(</span>dh <span class="token operator">+</span> <span class="token number">0.1</span><span class="token punctuation">)</span><span class="token punctuation">)</span>
left<span class="token punctuation">,</span> right <span class="token operator">=</span> <span class="token builtin">int</span><span class="token punctuation">(</span><span class="token builtin">round</span><span class="token punctuation">(</span>dw <span class="token operator">-</span> <span class="token number">0.1</span><span class="token punctuation">)</span><span class="token punctuation">)</span><span class="token punctuation">,</span> <span class="token builtin">int</span><span class="token punctuation">(</span><span class="token builtin">round</span><span class="token punctuation">(</span>dw <span class="token operator">+</span> <span class="token number">0.1</span><span class="token punctuation">)</span><span class="token punctuation">)</span>
im <span class="token operator">=</span> cv2<span class="token punctuation">.</span>copyMakeBorder<span class="token punctuation">(</span>im<span class="token punctuation">,</span> top<span class="token punctuation">,</span> bottom<span class="token punctuation">,</span> left<span class="token punctuation">,</span> right<span class="token punctuation">,</span> cv2<span class="token punctuation">.</span>BORDER_CONSTANT<span class="token punctuation">,</span> value<span class="token operator">=</span>pad_color<span class="token punctuation">)</span>  <span class="token comment"># add border</span>

<span class="token keyword">if</span> info_need <span class="token keyword">is</span> <span class="token boolean">True</span><span class="token punctuation">:</span>
    <span class="token keyword">return</span> im<span class="token punctuation">,</span> ratio<span class="token punctuation">,</span> <span class="token punctuation">(</span>dw<span class="token punctuation">,</span> dh<span class="token punctuation">)</span>
<span class="token keyword">else</span><span class="token punctuation">:</span>
    <span class="token keyword">return</span> im

def filter_boxes(boxes, box_confidences, box_class_probs):
“”“Filter boxes with object threshold.
“””
box_confidences = box_confidences.reshape(-1)
candidate, class_num = box_class_probs.shape

class_max_score <span class="token operator">=</span> np<span class="token punctuation">.</span><span class="token builtin">max</span><span class="token punctuation">(</span>box_class_probs<span class="token punctuation">,</span> axis<span class="token operator">=</span><span class="token operator">-</span><span class="token number">1</span><span class="token punctuation">)</span>
classes <span class="token operator">=</span> np<span class="token punctuation">.</span>argmax<span class="token punctuation">(</span>box_class_probs<span class="token punctuation">,</span> axis<span class="token operator">=</span><span class="token operator">-</span><span class="token number">1</span><span class="token punctuation">)</span>

_class_pos <span class="token operator">=</span> np<span class="token punctuation">.</span>where<span class="token punctuation">(</span>class_max_score <span class="token operator">*</span> box_confidences <span class="token operator">&gt;=</span> OBJ_THRESH<span class="token punctuation">)</span>
scores <span class="token operator">=</span> <span class="token punctuation">(</span>class_max_score <span class="token operator">*</span> box_confidences<span class="token punctuation">)</span><span class="token punctuation">[</span>_class_pos<span class="token punctuation">]</span>

boxes <span class="token operator">=</span> boxes<span class="token punctuation">[</span>_class_pos<span class="token punctuation">]</span>
classes <span class="token operator">=</span> classes<span class="token punctuation">[</span>_class_pos<span class="token punctuation">]</span>

<span class="token keyword">return</span> boxes<span class="token punctuation">,</span> classes<span class="token punctuation">,</span> scores

def nms_boxes(boxes, scores):
“”“Suppress non-maximal boxes.
# Returns
keep: ndarray, index of effective boxes.
“””
x = boxes[:, 0]
y = boxes[:, 1]
w = boxes[:, 2] - boxes[:, 0]
h = boxes[:, 3] - boxes[:, 1]

areas <span class="token operator">=</span> w <span class="token operator">*</span> h
order <span class="token operator">=</span> scores<span class="token punctuation">.</span>argsort<span class="token punctuation">(</span><span class="token punctuation">)</span><span class="token punctuation">[</span><span class="token punctuation">:</span><span class="token punctuation">:</span><span class="token operator">-</span><span class="token number">1</span><span class="token punctuation">]</span>

keep <span class="token operator">=</span> <span class="token punctuation">[</span><span class="token punctuation">]</span>
<span class="token keyword">while</span> order<span class="token punctuation">.</span>size <span class="token operator">&gt;</span> <span class="token number">0</span><span class="token punctuation">:</span>
    i <span class="token operator">=</span> order<span class="token punctuation">[</span><span class="token number">0</span><span class="token punctuation">]</span>
    keep<span class="token punctuation">.</span>append<span class="token punctuation">(</span>i<span class="token punctuation">)</span>

    xx1 <span class="token operator">=</span> np<span class="token punctuation">.</span>maximum<span class="token punctuation">(</span>x<span class="token punctuation">[</span>i<span class="token punctuation">]</span><span class="token punctuation">,</span> x<span class="token punctuation">[</span>order<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">:</span><span class="token punctuation">]</span><span class="token punctuation">]</span><span class="token punctuation">)</span>
    yy1 <span class="token operator">=</span> np<span class="token punctuation">.</span>maximum<span class="token punctuation">(</span>y<span class="token punctuation">[</span>i<span class="token punctuation">]</span><span class="token punctuation">,</span> y<span class="token punctuation">[</span>order<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">:</span><span class="token punctuation">]</span><span class="token punctuation">]</span><span class="token punctuation">)</span>
    xx2 <span class="token operator">=</span> np<span class="token punctuation">.</span>minimum<span class="token punctuation">(</span>x<span class="token punctuation">[</span>i<span class="token punctuation">]</span> <span class="token operator">+</span> w<span class="token punctuation">[</span>i<span class="token punctuation">]</span><span class="token punctuation">,</span> x<span class="token punctuation">[</span>order<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">:</span><span class="token punctuation">]</span><span class="token punctuation">]</span> <span class="token operator">+</span> w<span class="token punctuation">[</span>order<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">:</span><span class="token punctuation">]</span><span class="token punctuation">]</span><span class="token punctuation">)</span>
    yy2 <span class="token operator">=</span> np<span class="token punctuation">.</span>minimum<span class="token punctuation">(</span>y<span class="token punctuation">[</span>i<span class="token punctuation">]</span> <span class="token operator">+</span> h<span class="token punctuation">[</span>i<span class="token punctuation">]</span><span class="token punctuation">,</span> y<span class="token punctuation">[</span>order<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">:</span><span class="token punctuation">]</span><span class="token punctuation">]</span> <span class="token operator">+</span> h<span class="token punctuation">[</span>order<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">:</span><span class="token punctuation">]</span><span class="token punctuation">]</span><span class="token punctuation">)</span>

    w1 <span class="token operator">=</span> np<span class="token punctuation">.</span>maximum<span class="token punctuation">(</span><span class="token number">0.0</span><span class="token punctuation">,</span> xx2 <span class="token operator">-</span> xx1 <span class="token operator">+</span> <span class="token number">0.00001</span><span class="token punctuation">)</span>
    h1 <span class="token operator">=</span> np<span class="token punctuation">.</span>maximum<span class="token punctuation">(</span><span class="token number">0.0</span><span class="token punctuation">,</span> yy2 <span class="token operator">-</span> yy1 <span class="token operator">+</span> <span class="token number">0.00001</span><span class="token punctuation">)</span>
    inter <span class="token operator">=</span> w1 <span class="token operator">*</span> h1

    ovr <span class="token operator">=</span> inter <span class="token operator">/</span> <span class="token punctuation">(</span>areas<span class="token punctuation">[</span>i<span class="token punctuation">]</span> <span class="token operator">+</span> areas<span class="token punctuation">[</span>order<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">:</span><span class="token punctuation">]</span><span class="token punctuation">]</span> <span class="token operator">-</span> inter<span class="token punctuation">)</span>
    inds <span class="token operator">=</span> np<span class="token punctuation">.</span>where<span class="token punctuation">(</span>ovr <span class="token operator">&lt;=</span> NMS_THRESH<span class="token punctuation">)</span><span class="token punctuation">[</span><span class="token number">0</span><span class="token punctuation">]</span>
    order <span class="token operator">=</span> order<span class="token punctuation">[</span>inds <span class="token operator">+</span> <span class="token number">1</span><span class="token punctuation">]</span>
keep <span class="token operator">=</span> np<span class="token punctuation">.</span>array<span class="token punctuation">(</span>keep<span class="token punctuation">)</span>
<span class="token keyword">return</span> keep

def softmax(x, axis=None):
x = x - x.max(axis=axis, keepdims=True)
y = np.exp(x)
return y / y.sum(axis=axis, keepdims=True)

def dfl(position):
# Distribution Focal Loss (DFL)
n, c, h, w = position.shape
p_num = 4
mc = c // p_num
y = position.reshape(n, p_num, mc, h, w)
y = softmax(y, 2)
acc_metrix = np.array(range(mc), dtype=float).reshape(1, 1, mc, 1, 1)
y = (y * acc_metrix).sum(2)
return y

def box_process(position):
grid_h, grid_w = position.shape[2:4]
col, row = np.meshgrid(np.arange(0, grid_w), np.arange(0, grid_h))
col = col.reshape(1, 1, grid_h, grid_w)
row = row.reshape(1, 1, grid_h, grid_w)
grid = np.concatenate((col, row), axis=1)
stride = np.array([MODEL_SIZE[1] // grid_h, MODEL_SIZE[0] // grid_w]).reshape(1, 2, 1, 1)

position <span class="token operator">=</span> dfl<span class="token punctuation">(</span>position<span class="token punctuation">)</span>
box_xy <span class="token operator">=</span> grid <span class="token operator">+</span> <span class="token number">0.5</span> <span class="token operator">-</span> position<span class="token punctuation">[</span><span class="token punctuation">:</span><span class="token punctuation">,</span> <span class="token number">0</span><span class="token punctuation">:</span><span class="token number">2</span><span class="token punctuation">,</span> <span class="token punctuation">:</span><span class="token punctuation">,</span> <span class="token punctuation">:</span><span class="token punctuation">]</span>
box_xy2 <span class="token operator">=</span> grid <span class="token operator">+</span> <span class="token number">0.5</span> <span class="token operator">+</span> position<span class="token punctuation">[</span><span class="token punctuation">:</span><span class="token punctuation">,</span> <span class="token number">2</span><span class="token punctuation">:</span><span class="token number">4</span><span class="token punctuation">,</span> <span class="token punctuation">:</span><span class="token punctuation">,</span> <span class="token punctuation">:</span><span class="token punctuation">]</span>
xyxy <span class="token operator">=</span> np<span class="token punctuation">.</span>concatenate<span class="token punctuation">(</span><span class="token punctuation">(</span>box_xy <span class="token operator">*</span> stride<span class="token punctuation">,</span> box_xy2 <span class="token operator">*</span> stride<span class="token punctuation">)</span><span class="token punctuation">,</span> axis<span class="token operator">=</span><span class="token number">1</span><span class="token punctuation">)</span>

<span class="token keyword">return</span> xyxy

def post_process(input_data):
boxes, scores, classes_conf = [], [], []
defualt_branch = 3
pair_per_branch = len(input_data) // defualt_branch
# Python 忽略 score_sum 输出
for i in range(defualt_branch):
boxes.append(box_process(input_data[pair_per_branch i]))
classes_conf.append(input_data[pair_per_branch i + 1])
scores.append(np.ones_like(input_data[pair_per_branch * i + 1][:, :1, :, :], dtype=np.float32))

<span class="token keyword">def</span> <span class="token function">sp_flatten</span><span class="token punctuation">(</span>_in<span class="token punctuation">)</span><span class="token punctuation">:</span>
    ch <span class="token operator">=</span> _in<span class="token punctuation">.</span>shape<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">]</span>
    _in <span class="token operator">=</span> _in<span class="token punctuation">.</span>transpose<span class="token punctuation">(</span><span class="token number">0</span><span class="token punctuation">,</span> <span class="token number">2</span><span class="token punctuation">,</span> <span class="token number">3</span><span class="token punctuation">,</span> <span class="token number">1</span><span class="token punctuation">)</span>
    <span class="token keyword">return</span> _in<span class="token punctuation">.</span>reshape<span class="token punctuation">(</span><span class="token operator">-</span><span class="token number">1</span><span class="token punctuation">,</span> ch<span class="token punctuation">)</span>

boxes <span class="token operator">=</span> <span class="token punctuation">[</span>sp_flatten<span class="token punctuation">(</span>_v<span class="token punctuation">)</span> <span class="token keyword">for</span> _v <span class="token keyword">in</span> boxes<span class="token punctuation">]</span>
classes_conf <span class="token operator">=</span> <span class="token punctuation">[</span>sp_flatten<span class="token punctuation">(</span>_v<span class="token punctuation">)</span> <span class="token keyword">for</span> _v <span class="token keyword">in</span> classes_conf<span class="token punctuation">]</span>
scores <span class="token operator">=</span> <span class="token punctuation">[</span>sp_flatten<span class="token punctuation">(</span>_v<span class="token punctuation">)</span> <span class="token keyword">for</span> _v <span class="token keyword">in</span> scores<span class="token punctuation">]</span>

boxes <span class="token operator">=</span> np<span class="token punctuation">.</span>concatenate<span class="token punctuation">(</span>boxes<span class="token punctuation">)</span>
classes_conf <span class="token operator">=</span> np<span class="token punctuation">.</span>concatenate<span class="token punctuation">(</span>classes_conf<span class="token punctuation">)</span>
scores <span class="token operator">=</span> np<span class="token punctuation">.</span>concatenate<span class="token punctuation">(</span>scores<span class="token punctuation">)</span>

<span class="token comment"># filter according to threshold</span>
boxes<span class="token punctuation">,</span> classes<span class="token punctuation">,</span> scores <span class="token operator">=</span> filter_boxes<span class="token punctuation">(</span>boxes<span class="token punctuation">,</span> scores<span class="token punctuation">,</span> classes_conf<span class="token punctuation">)</span>

<span class="token comment"># nms</span>
nboxes<span class="token punctuation">,</span> nclasses<span class="token punctuation">,</span> nscores <span class="token operator">=</span> <span class="token punctuation">[</span><span class="token punctuation">]</span><span class="token punctuation">,</span> <span class="token punctuation">[</span><span class="token punctuation">]</span><span class="token punctuation">,</span> <span class="token punctuation">[</span><span class="token punctuation">]</span>
<span class="token keyword">for</span> c <span class="token keyword">in</span> <span class="token builtin">set</span><span class="token punctuation">(</span>classes<span class="token punctuation">)</span><span class="token punctuation">:</span>
    inds <span class="token operator">=</span> np<span class="token punctuation">.</span>where<span class="token punctuation">(</span>classes <span class="token operator">==</span> c<span class="token punctuation">)</span>
    b <span class="token operator">=</span> boxes<span class="token punctuation">[</span>inds<span class="token punctuation">]</span>
    c <span class="token operator">=</span> classes<span class="token punctuation">[</span>inds<span class="token punctuation">]</span>
    s <span class="token operator">=</span> scores<span class="token punctuation">[</span>inds<span class="token punctuation">]</span>
    keep <span class="token operator">=</span> nms_boxes<span class="token punctuation">(</span>b<span class="token punctuation">,</span> s<span class="token punctuation">)</span>

    <span class="token keyword">if</span> <span class="token builtin">len</span><span class="token punctuation">(</span>keep<span class="token punctuation">)</span> <span class="token operator">!=</span> <span class="token number">0</span><span class="token punctuation">:</span>
        nboxes<span class="token punctuation">.</span>append<span class="token punctuation">(</span>b<span class="token punctuation">[</span>keep<span class="token punctuation">]</span><span class="token punctuation">)</span>
        nclasses<span class="token punctuation">.</span>append<span class="token punctuation">(</span>c<span class="token punctuation">[</span>keep<span class="token punctuation">]</span><span class="token punctuation">)</span>
        nscores<span class="token punctuation">.</span>append<span class="token punctuation">(</span>s<span class="token punctuation">[</span>keep<span class="token punctuation">]</span><span class="token punctuation">)</span>

<span class="token keyword">if</span> <span class="token keyword">not</span> nclasses <span class="token keyword">and</span> <span class="token keyword">not</span> nscores<span class="token punctuation">:</span>
    <span class="token keyword">return</span> <span class="token boolean">None</span><span class="token punctuation">,</span> <span class="token boolean">None</span><span class="token punctuation">,</span> <span class="token boolean">None</span>

boxes <span class="token operator">=</span> np<span class="token punctuation">.</span>concatenate<span class="token punctuation">(</span>nboxes<span class="token punctuation">)</span>
classes <span class="token operator">=</span> np<span class="token punctuation">.</span>concatenate<span class="token punctuation">(</span>nclasses<span class="token punctuation">)</span>
scores <span class="token operator">=</span> np<span class="token punctuation">.</span>concatenate<span class="token punctuation">(</span>nscores<span class="token punctuation">)</span>

<span class="token keyword">return</span> boxes<span class="token punctuation">,</span> classes<span class="token punctuation">,</span> scores

def draw_detections(img, left, top, right, bottom, score, class_id):
“”"
Draws bounding boxes and labels on the input image based on the detected objects.
Args:
img: The input image to draw detections on.
box: Detected bounding box.
score: Corresponding detection score.
class_id: Class ID for the detected object.
Returns:
None
“”"

<span class="token comment"># Retrieve the color for the class ID</span>
color <span class="token operator">=</span> color_palette<span class="token punctuation">[</span>class_id<span class="token punctuation">]</span>

<span class="token comment"># Draw the bounding box on the image</span>
cv2<span class="token punctuation">.</span>rectangle<span class="token punctuation">(</span>img<span class="token punctuation">,</span> <span class="token punctuation">(</span><span class="token builtin">int</span><span class="token punctuation">(</span>left<span class="token punctuation">)</span><span class="token punctuation">,</span> <span class="token builtin">int</span><span class="token punctuation">(</span>top<span class="token punctuation">)</span><span class="token punctuation">)</span><span class="token punctuation">,</span> <span class="token punctuation">(</span><span class="token builtin">int</span><span class="token punctuation">(</span>right<span class="token punctuation">)</span><span class="token punctuation">,</span> <span class="token builtin">int</span><span class="token punctuation">(</span>bottom<span class="token punctuation">)</span><span class="token punctuation">)</span><span class="token punctuation">,</span> color<span class="token punctuation">,</span> <span class="token number">2</span><span class="token punctuation">)</span>

<span class="token comment"># Create the label text with class name and score</span>
label <span class="token operator">=</span> <span class="token string-interpolation"><span class="token string">f"</span><span class="token interpolation"><span class="token punctuation">{<!-- --></span>CLASSES<span class="token punctuation">[</span>class_id<span class="token punctuation">]</span><span class="token punctuation">}</span></span><span class="token string">: </span><span class="token interpolation"><span class="token punctuation">{<!-- --></span>score<span class="token punctuation">:</span><span class="token format-spec">.2f</span><span class="token punctuation">}</span></span><span class="token string">"</span></span>

<span class="token comment"># Calculate the dimensions of the label text</span>
<span class="token punctuation">(</span>label_width<span class="token punctuation">,</span> label_height<span class="token punctuation">)</span><span class="token punctuation">,</span> _ <span class="token operator">=</span> cv2<span class="token punctuation">.</span>getTextSize<span class="token punctuation">(</span>label<span class="token punctuation">,</span> cv2<span class="token punctuation">.</span>FONT_HERSHEY_SIMPLEX<span class="token punctuation">,</span> <span class="token number">0.5</span><span class="token punctuation">,</span> <span class="token number">1</span><span class="token punctuation">)</span>

<span class="token comment"># Calculate the position of the label text</span>
label_x <span class="token operator">=</span> left
label_y <span class="token operator">=</span> top <span class="token operator">-</span> <span class="token number">10</span> <span class="token keyword">if</span> top <span class="token operator">-</span> <span class="token number">10</span> <span class="token operator">&gt;</span> label_height <span class="token keyword">else</span> top <span class="token operator">+</span> <span class="token number">10</span>

<span class="token comment"># Draw a filled rectangle as the background for the label text</span>
cv2<span class="token punctuation">.</span>rectangle<span class="token punctuation">(</span>img<span class="token punctuation">,</span> <span class="token punctuation">(</span>label_x<span class="token punctuation">,</span> label_y <span class="token operator">-</span> label_height<span class="token punctuation">)</span><span class="token punctuation">,</span> <span class="token punctuation">(</span>label_x <span class="token operator">+</span> label_width<span class="token punctuation">,</span> label_y <span class="token operator">+</span> label_height<span class="token punctuation">)</span><span class="token punctuation">,</span> color<span class="token punctuation">,</span>
              cv2<span class="token punctuation">.</span>FILLED<span class="token punctuation">)</span>

<span class="token comment"># Draw the label text on the image</span>
cv2<span class="token punctuation">.</span>putText<span class="token punctuation">(</span>img<span class="token punctuation">,</span> label<span class="token punctuation">,</span> <span class="token punctuation">(</span>label_x<span class="token punctuation">,</span> label_y<span class="token punctuation">)</span><span class="token punctuation">,</span> cv2<span class="token punctuation">.</span>FONT_HERSHEY_SIMPLEX<span class="token punctuation">,</span> <span class="token number">0.5</span><span class="token punctuation">,</span> <span class="token punctuation">(</span><span class="token number">0</span><span class="token punctuation">,</span> <span class="token number">0</span><span class="token punctuation">,</span> <span class="token number">0</span><span class="token punctuation">)</span><span class="token punctuation">,</span> <span class="token number">1</span><span class="token punctuation">,</span> cv2<span class="token punctuation">.</span>LINE_AA<span class="token punctuation">)</span>

def draw(image, boxes, scores, classes):
img_h, img_w = image.shape[:2]
# Calculate scaling factors for bounding box coordinates
x_factor = img_w / MODEL_SIZE[0]
y_factor = img_h / MODEL_SIZE[1]

<span class="token keyword">for</span> box<span class="token punctuation">,</span> score<span class="token punctuation">,</span> cl <span class="token keyword">in</span> <span class="token builtin">zip</span><span class="token punctuation">(</span>boxes<span class="token punctuation">,</span> scores<span class="token punctuation">,</span> classes<span class="token punctuation">)</span><span class="token punctuation">:</span>
    x1<span class="token punctuation">,</span> y1<span class="token punctuation">,</span> x2<span class="token punctuation">,</span> y2 <span class="token operator">=</span> <span class="token punctuation">[</span><span class="token builtin">int</span><span class="token punctuation">(</span>_b<span class="token punctuation">)</span> <span class="token keyword">for</span> _b <span class="token keyword">in</span> box<span class="token punctuation">]</span>

    left <span class="token operator">=</span> <span class="token builtin">int</span><span class="token punctuation">(</span>x1 <span class="token operator">*</span> x_factor<span class="token punctuation">)</span>
    top <span class="token operator">=</span> <span class="token builtin">int</span><span class="token punctuation">(</span>y1 <span class="token operator">*</span> y_factor<span class="token punctuation">)</span> <span class="token operator">-</span> <span class="token number">10</span>
    right <span class="token operator">=</span> <span class="token builtin">int</span><span class="token punctuation">(</span>x2 <span class="token operator">*</span> x_factor<span class="token punctuation">)</span>
    bottom <span class="token operator">=</span> <span class="token builtin">int</span><span class="token punctuation">(</span>y2 <span class="token operator">*</span> y_factor<span class="token punctuation">)</span> <span class="token operator">+</span> <span class="token number">10</span>

    <span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'class: {}, score: {}'</span><span class="token punctuation">.</span><span class="token builtin">format</span><span class="token punctuation">(</span>CLASSES<span class="token punctuation">[</span>cl<span class="token punctuation">]</span><span class="token punctuation">,</span> score<span class="token punctuation">)</span><span class="token punctuation">)</span>
    <span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'box coordinate left,top,right,down: [{}, {}, {}, {}]'</span><span class="token punctuation">.</span><span class="token builtin">format</span><span class="token punctuation">(</span>left<span class="token punctuation">,</span> top<span class="token punctuation">,</span> right<span class="token punctuation">,</span> bottom<span class="token punctuation">)</span><span class="token punctuation">)</span>

    <span class="token comment"># Retrieve the color for the class ID</span>

    draw_detections<span class="token punctuation">(</span>image<span class="token punctuation">,</span> left<span class="token punctuation">,</span> top<span class="token punctuation">,</span> right<span class="token punctuation">,</span> bottom<span class="token punctuation">,</span> score<span class="token punctuation">,</span> cl<span class="token punctuation">)</span>

    <span class="token comment"># cv2.rectangle(image, (left, top), (right, bottom), color, 2)</span>
    <span class="token comment"># cv2.putText(image, '{0} {1:.2f}'.format(CLASSES[cl], score),</span>
    <span class="token comment">#             (left, top - 6),</span>
    <span class="token comment">#             cv2.FONT_HERSHEY_SIMPLEX,</span>
    <span class="token comment">#             0.6, (0, 0, 255), 2)</span>

if name == ‘main’:

<span class="token comment"># 确定目标设备</span>
target <span class="token operator">=</span> <span class="token string">'RK3588'</span>
<span class="token comment"># 创建RKNN对象</span>
rknn <span class="token operator">=</span> RKNN<span class="token punctuation">(</span><span class="token punctuation">)</span>

<span class="token comment"># 配置RKNN模型</span>
<span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'--&gt; config model'</span><span class="token punctuation">)</span>
rknn<span class="token punctuation">.</span>config<span class="token punctuation">(</span>
    mean_values<span class="token operator">=</span><span class="token punctuation">[</span><span class="token punctuation">[</span><span class="token number">0</span><span class="token punctuation">,</span> <span class="token number">0</span><span class="token punctuation">,</span> <span class="token number">0</span><span class="token punctuation">]</span><span class="token punctuation">]</span><span class="token punctuation">,</span>
    std_values<span class="token operator">=</span><span class="token punctuation">[</span><span class="token punctuation">[</span><span class="token number">255</span><span class="token punctuation">,</span> <span class="token number">255</span><span class="token punctuation">,</span> <span class="token number">255</span><span class="token punctuation">]</span><span class="token punctuation">]</span><span class="token punctuation">,</span>
    target_platform<span class="token operator">=</span>target<span class="token punctuation">,</span>
<span class="token punctuation">)</span>
<span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'done'</span><span class="token punctuation">)</span>

<span class="token comment"># 加载 .onnx模型</span>
<span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'--&gt; loading model'</span><span class="token punctuation">)</span>
ret <span class="token operator">=</span> rknn<span class="token punctuation">.</span>load_onnx<span class="token punctuation">(</span>model<span class="token operator">=</span><span class="token string">"./best.onnx"</span><span class="token punctuation">)</span>
<span class="token keyword">if</span> ret <span class="token operator">!=</span> <span class="token number">0</span><span class="token punctuation">:</span>
    <span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">"load model failed!"</span><span class="token punctuation">)</span>
    rknn<span class="token punctuation">.</span>release<span class="token punctuation">(</span><span class="token punctuation">)</span>
    exit<span class="token punctuation">(</span>ret<span class="token punctuation">)</span>
<span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'done'</span><span class="token punctuation">)</span>

<span class="token comment"># 构建RKNN模型</span>
<span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'--&gt; building model'</span><span class="token punctuation">)</span>
ret <span class="token operator">=</span> rknn<span class="token punctuation">.</span>build<span class="token punctuation">(</span>do_quantization<span class="token operator">=</span><span class="token boolean">True</span><span class="token punctuation">,</span> dataset<span class="token operator">=</span><span class="token string">"./dataset/dataset.txt"</span><span class="token punctuation">)</span>
<span class="token keyword">if</span> ret <span class="token operator">!=</span> <span class="token number">0</span><span class="token punctuation">:</span>
    <span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">"build model failed!"</span><span class="token punctuation">)</span>
    rknn<span class="token punctuation">.</span>release<span class="token punctuation">(</span><span class="token punctuation">)</span>
    exit<span class="token punctuation">(</span>ret<span class="token punctuation">)</span>
<span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'done'</span><span class="token punctuation">)</span>

<span class="token comment"># 导出RKNN模型</span>
<span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'--&gt;export RKNN model'</span><span class="token punctuation">)</span>
ret <span class="token operator">=</span> rknn<span class="token punctuation">.</span>export_rknn<span class="token punctuation">(</span><span class="token string">'./yolov8.rknn'</span><span class="token punctuation">)</span>
<span class="token keyword">if</span> ret <span class="token operator">!=</span> <span class="token number">0</span><span class="token punctuation">:</span>
    <span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'export RKNN model failed'</span><span class="token punctuation">)</span>
    rknn<span class="token punctuation">.</span>release<span class="token punctuation">(</span><span class="token punctuation">)</span>
    exit<span class="token punctuation">(</span>ret<span class="token punctuation">)</span>

<span class="token comment"># 初始化 runtime 环境</span>
<span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'--&gt; Init runtime environment'</span><span class="token punctuation">)</span>
<span class="token comment"># run on RK356x/RK3588 with Debian OS, do not need specify target.</span>
<span class="token comment">#ret = rknn.init_runtime(target='rk3588', device_id='48c122b87375ccbc')</span>
<span class="token comment"># 如果使用电脑进行模拟测试</span>
ret <span class="token operator">=</span> rknn<span class="token punctuation">.</span>init_runtime<span class="token punctuation">(</span><span class="token punctuation">)</span>

<span class="token keyword">if</span> ret <span class="token operator">!=</span> <span class="token number">0</span><span class="token punctuation">:</span>
    <span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'Init runtime environment failed!'</span><span class="token punctuation">)</span>
    exit<span class="token punctuation">(</span>ret<span class="token punctuation">)</span>
<span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'done'</span><span class="token punctuation">)</span>

<span class="token comment"># 数据处理</span>
img_list <span class="token operator">=</span> os<span class="token punctuation">.</span>listdir<span class="token punctuation">(</span>IMG_FOLDER<span class="token punctuation">)</span>
<span class="token keyword">for</span> i <span class="token keyword">in</span> <span class="token builtin">range</span><span class="token punctuation">(</span><span class="token builtin">len</span><span class="token punctuation">(</span>img_list<span class="token punctuation">)</span><span class="token punctuation">)</span><span class="token punctuation">:</span>
    img_name <span class="token operator">=</span> img_list<span class="token punctuation">[</span>i<span class="token punctuation">]</span>
    img_path <span class="token operator">=</span> os<span class="token punctuation">.</span>path<span class="token punctuation">.</span>join<span class="token punctuation">(</span>IMG_FOLDER<span class="token punctuation">,</span> img_name<span class="token punctuation">)</span>
    <span class="token keyword">if</span> <span class="token keyword">not</span> os<span class="token punctuation">.</span>path<span class="token punctuation">.</span>exists<span class="token punctuation">(</span>img_path<span class="token punctuation">)</span><span class="token punctuation">:</span>
        <span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">"{} is not found"</span><span class="token punctuation">,</span> img_name<span class="token punctuation">)</span>
        <span class="token keyword">continue</span>
    img_src <span class="token operator">=</span> cv2<span class="token punctuation">.</span>imread<span class="token punctuation">(</span>img_path<span class="token punctuation">)</span>
    <span class="token keyword">if</span> img_src <span class="token keyword">is</span> <span class="token boolean">None</span><span class="token punctuation">:</span>
        <span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">"文件不存在\n"</span><span class="token punctuation">)</span>


    <span class="token comment"># Due to rga init with (0,0,0), we using pad_color (0,0,0) instead of (114, 114, 114)</span>
pad_color <span class="token operator">=</span> <span class="token punctuation">(</span><span class="token number">0</span><span class="token punctuation">,</span> <span class="token number">0</span><span class="token punctuation">,</span> <span class="token number">0</span><span class="token punctuation">)</span>
img <span class="token operator">=</span> letter_box<span class="token punctuation">(</span>im<span class="token operator">=</span>img_src<span class="token punctuation">.</span>copy<span class="token punctuation">(</span><span class="token punctuation">)</span><span class="token punctuation">,</span> new_shape<span class="token operator">=</span><span class="token punctuation">(</span>MODEL_SIZE<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">]</span><span class="token punctuation">,</span> MODEL_SIZE<span class="token punctuation">[</span><span class="token number">0</span><span class="token punctuation">]</span><span class="token punctuation">)</span><span class="token punctuation">,</span> pad_color<span class="token operator">=</span><span class="token punctuation">(</span><span class="token number">0</span><span class="token punctuation">,</span> <span class="token number">0</span><span class="token punctuation">,</span> <span class="token number">0</span><span class="token punctuation">)</span><span class="token punctuation">)</span>
<span class="token comment"># img = cv2.resize(img_src, (640, 512), interpolation=cv2.INTER_LINEAR) # direct resize</span>
<span class="token builtin">input</span> <span class="token operator">=</span> np<span class="token punctuation">.</span>expand_dims<span class="token punctuation">(</span>img<span class="token punctuation">,</span> axis<span class="token operator">=</span><span class="token number">0</span><span class="token punctuation">)</span>

outputs <span class="token operator">=</span> rknn<span class="token punctuation">.</span>inference<span class="token punctuation">(</span><span class="token punctuation">[</span><span class="token builtin">input</span><span class="token punctuation">]</span><span class="token punctuation">)</span>

boxes<span class="token punctuation">,</span> classes<span class="token punctuation">,</span> scores <span class="token operator">=</span> post_process<span class="token punctuation">(</span>outputs<span class="token punctuation">)</span>

img_p <span class="token operator">=</span> img_src<span class="token punctuation">.</span>copy<span class="token punctuation">(</span><span class="token punctuation">)</span>

<span class="token keyword">if</span> boxes <span class="token keyword">is</span> <span class="token keyword">not</span> <span class="token boolean">None</span><span class="token punctuation">:</span>
    draw<span class="token punctuation">(</span>img_p<span class="token punctuation">,</span> boxes<span class="token punctuation">,</span> scores<span class="token punctuation">,</span> classes<span class="token punctuation">)</span>

<span class="token comment"># 保存结果</span>
<span class="token keyword">if</span> <span class="token keyword">not</span> os<span class="token punctuation">.</span>path<span class="token punctuation">.</span>exists<span class="token punctuation">(</span>RESULT_PATH<span class="token punctuation">)</span><span class="token punctuation">:</span>
    os<span class="token punctuation">.</span>mkdir<span class="token punctuation">(</span>RESULT_PATH<span class="token punctuation">)</span>

result_path <span class="token operator">=</span> os<span class="token punctuation">.</span>path<span class="token punctuation">.</span>join<span class="token punctuation">(</span>RESULT_PATH<span class="token punctuation">,</span> img_name<span class="token punctuation">)</span>
cv2<span class="token punctuation">.</span>imwrite<span class="token punctuation">(</span>result_path<span class="token punctuation">,</span> img_p<span class="token punctuation">)</span>
<span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'Detection result save to {}'</span><span class="token punctuation">.</span><span class="token builtin">format</span><span class="token punctuation">(</span>result_path<span class="token punctuation">)</span><span class="token punctuation">)</span>

<span class="token keyword">pass</span>

rknn<span class="token punctuation">.</span>release<span class="token punctuation">(</span><span class="token punctuation">)</span>
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15
  • 16
  • 17
  • 18
  • 19
  • 20
  • 21
  • 22
  • 23
  • 24
  • 25
  • 26
  • 27
  • 28
  • 29
  • 30
  • 31
  • 32
  • 33
  • 34
  • 35
  • 36
  • 37
  • 38
  • 39
  • 40
  • 41
  • 42
  • 43
  • 44
  • 45
  • 46
  • 47
  • 48
  • 49
  • 50
  • 51
  • 52
  • 53
  • 54
  • 55
  • 56
  • 57
  • 58
  • 59
  • 60
  • 61
  • 62
  • 63
  • 64
  • 65
  • 66
  • 67
  • 68
  • 69
  • 70
  • 71
  • 72
  • 73
  • 74
  • 75
  • 76
  • 77
  • 78
  • 79
  • 80
  • 81
  • 82
  • 83
  • 84
  • 85
  • 86
  • 87
  • 88
  • 89
  • 90
  • 91
  • 92
  • 93
  • 94
  • 95
  • 96
  • 97
  • 98
  • 99
  • 100
  • 101
  • 102
  • 103
  • 104
  • 105
  • 106
  • 107
  • 108
  • 109
  • 110
  • 111
  • 112
  • 113
  • 114
  • 115
  • 116
  • 117
  • 118
  • 119
  • 120
  • 121
  • 122
  • 123
  • 124
  • 125
  • 126
  • 127
  • 128
  • 129
  • 130
  • 131
  • 132
  • 133
  • 134
  • 135
  • 136
  • 137
  • 138
  • 139
  • 140
  • 141
  • 142
  • 143
  • 144
  • 145
  • 146
  • 147
  • 148
  • 149
  • 150
  • 151
  • 152
  • 153
  • 154
  • 155
  • 156
  • 157
  • 158
  • 159
  • 160
  • 161
  • 162
  • 163
  • 164
  • 165
  • 166
  • 167
  • 168
  • 169
  • 170
  • 171
  • 172
  • 173
  • 174
  • 175
  • 176
  • 177
  • 178
  • 179
  • 180
  • 181
  • 182
  • 183
  • 184
  • 185
  • 186
  • 187
  • 188
  • 189
  • 190
  • 191
  • 192
  • 193
  • 194
  • 195
  • 196
  • 197
  • 198
  • 199
  • 200
  • 201
  • 202
  • 203
  • 204
  • 205
  • 206
  • 207
  • 208
  • 209
  • 210
  • 211
  • 212
  • 213
  • 214
  • 215
  • 216
  • 217
  • 218
  • 219
  • 220
  • 221
  • 222
  • 223
  • 224
  • 225
  • 226
  • 227
  • 228
  • 229
  • 230
  • 231
  • 232
  • 233
  • 234
  • 235
  • 236
  • 237
  • 238
  • 239
  • 240
  • 241
  • 242
  • 243
  • 244
  • 245
  • 246
  • 247
  • 248
  • 249
  • 250
  • 251
  • 252
  • 253
  • 254
  • 255
  • 256
  • 257
  • 258
  • 259
  • 260
  • 261
  • 262
  • 263
  • 264
  • 265
  • 266
  • 267
  • 268
  • 269
  • 270
  • 271
  • 272
  • 273
  • 274
  • 275
  • 276
  • 277
  • 278
  • 279
  • 280
  • 281
  • 282
  • 283
  • 284
  • 285
  • 286
  • 287
  • 288
  • 289
  • 290
  • 291
  • 292
  • 293
  • 294
  • 295
  • 296
  • 297
  • 298
  • 299
  • 300
  • 301
  • 302
  • 303
  • 304
  • 305
  • 306
  • 307
  • 308
  • 309
  • 310
  • 311
  • 312
  • 313
  • 314
  • 315
  • 316
  • 317
  • 318
  • 319
  • 320
  • 321
  • 322
  • 323
  • 324
  • 325
  • 326
  • 327
  • 328
  • 329
  • 330
  • 331
  • 332
  • 333
  • 334
  • 335
  • 336
  • 337
  • 338
  • 339
  • 340
  • 341
  • 342
  • 343
  • 344
  • 345

运行inference_test-2.py 文件，会在yolov8_test文件夹下生成yolov8.rknn，测试结果的图片保存在dataset-2文件夹下。

python inference_test-2.py 

 
 

 
 
  • 1

五，模型部署

5.1 香橙派（RK3588）环境安装

首先参考Orange Pi - Orangepi官网安装Ubuntu系统

然后打开时ssh服务，我这里使用的MobaXterm进行连接的

sudo apt-get install openssh-server

 
 

 
 
  • 1

首先安装Miniconda — Anaconda documentation

外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传

下载后使用命令来安装Miniconda3

./Miniconda3-latest-Linux-aarch64.sh

 
 

 
 
  • 1

更新环境变量

source /home/topeet/.bashrc

 
 

 
 
  • 1

创建环境

conda create -n rknn python=3.9

 
 

 
 
  • 1

激活环境

conda activate rknn

 
 

 
 
  • 1

安装rknn_toolkit_lite2

在4.1下载的rknn-toolkit2-1.6.0/rknn_toolkit_lite2/packages中有rknn_toolkit_lite2-1.6.0-cp39-cp39-linux_aarch64.whl，把它移动到开发板上然后使用下面的命令安装

pip install rknn_toolkit_lite2-1.6.0-cp39-cp39-linux_aarch64.whl -i https://pypi.mirrors.ustc.edu.cn/simple/

 
 

 
 
  • 1

安装opencv

pip install opencv-python -i https://pypi.tuna.tsinghua.edu.cn/simple

 
 

 
 
  • 1

5.2 python部署

在开发板上新建一个yolov8_test来存放文件，dataset-1是存放的测试图片，dataset-2是保存测试结果的位置，yolov8.rknn是我们的rknn模型，rk3588_test-2.py是我们的测试脚本

.
├── dataset-1
│   └── 000000000650.jpg
├── dataset-2
├── rk3588_test-2.py
└── yolov8.rknn

 
 

 
 
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6

rk3588_test-2.py文件如下所示

import os
import cv2
from rknnlite.api import RKNNLite
import numpy as np

RKNN_MODEL = “./yolov8.rknn”
IMG_FOLDER = “dataset-1”
RESULT_PATH = ‘./dataset-2’

CLASSES = [‘car’]

OBJ_THRESH = 0.45
NMS_THRESH = 0.45

MODEL_SIZE = (640, 640)

color_palette = np.random.uniform(0, 255, size=(len(CLASSES), 3))

def sigmoid(x):
return 1 / (1 + np.exp(-x))

def letter_box(im, new_shape, pad_color=(0, 0, 0), info_need=False):
# Resize and pad image while meeting stride-multiple constraints
shape = im.shape[:2] # current shape [height, width]
if isinstance(new_shape, int):
new_shape = (new_shape, new_shape)

<span class="token comment"># Scale ratio</span>
r <span class="token operator">=</span> <span class="token builtin">min</span><span class="token punctuation">(</span>new_shape<span class="token punctuation">[</span><span class="token number">0</span><span class="token punctuation">]</span> <span class="token operator">/</span> shape<span class="token punctuation">[</span><span class="token number">0</span><span class="token punctuation">]</span><span class="token punctuation">,</span> new_shape<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">]</span> <span class="token operator">/</span> shape<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">]</span><span class="token punctuation">)</span>

<span class="token comment"># Compute padding</span>
ratio <span class="token operator">=</span> r  <span class="token comment"># width, height ratios</span>
new_unpad <span class="token operator">=</span> <span class="token builtin">int</span><span class="token punctuation">(</span><span class="token builtin">round</span><span class="token punctuation">(</span>shape<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">]</span> <span class="token operator">*</span> r<span class="token punctuation">)</span><span class="token punctuation">)</span><span class="token punctuation">,</span> <span class="token builtin">int</span><span class="token punctuation">(</span><span class="token builtin">round</span><span class="token punctuation">(</span>shape<span class="token punctuation">[</span><span class="token number">0</span><span class="token punctuation">]</span> <span class="token operator">*</span> r<span class="token punctuation">)</span><span class="token punctuation">)</span>
dw<span class="token punctuation">,</span> dh <span class="token operator">=</span> new_shape<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">]</span> <span class="token operator">-</span> new_unpad<span class="token punctuation">[</span><span class="token number">0</span><span class="token punctuation">]</span><span class="token punctuation">,</span> new_shape<span class="token punctuation">[</span><span class="token number">0</span><span class="token punctuation">]</span> <span class="token operator">-</span> new_unpad<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">]</span>  <span class="token comment"># wh padding</span>

dw <span class="token operator">/=</span> <span class="token number">2</span>  <span class="token comment"># divide padding into 2 sides</span>
dh <span class="token operator">/=</span> <span class="token number">2</span>

<span class="token keyword">if</span> shape<span class="token punctuation">[</span><span class="token punctuation">:</span><span class="token punctuation">:</span><span class="token operator">-</span><span class="token number">1</span><span class="token punctuation">]</span> <span class="token operator">!=</span> new_unpad<span class="token punctuation">:</span>  <span class="token comment"># resize</span>
    im <span class="token operator">=</span> cv2<span class="token punctuation">.</span>resize<span class="token punctuation">(</span>im<span class="token punctuation">,</span> new_unpad<span class="token punctuation">,</span> interpolation<span class="token operator">=</span>cv2<span class="token punctuation">.</span>INTER_LINEAR<span class="token punctuation">)</span>
top<span class="token punctuation">,</span> bottom <span class="token operator">=</span> <span class="token builtin">int</span><span class="token punctuation">(</span><span class="token builtin">round</span><span class="token punctuation">(</span>dh <span class="token operator">-</span> <span class="token number">0.1</span><span class="token punctuation">)</span><span class="token punctuation">)</span><span class="token punctuation">,</span> <span class="token builtin">int</span><span class="token punctuation">(</span><span class="token builtin">round</span><span class="token punctuation">(</span>dh <span class="token operator">+</span> <span class="token number">0.1</span><span class="token punctuation">)</span><span class="token punctuation">)</span>
left<span class="token punctuation">,</span> right <span class="token operator">=</span> <span class="token builtin">int</span><span class="token punctuation">(</span><span class="token builtin">round</span><span class="token punctuation">(</span>dw <span class="token operator">-</span> <span class="token number">0.1</span><span class="token punctuation">)</span><span class="token punctuation">)</span><span class="token punctuation">,</span> <span class="token builtin">int</span><span class="token punctuation">(</span><span class="token builtin">round</span><span class="token punctuation">(</span>dw <span class="token operator">+</span> <span class="token number">0.1</span><span class="token punctuation">)</span><span class="token punctuation">)</span>
im <span class="token operator">=</span> cv2<span class="token punctuation">.</span>copyMakeBorder<span class="token punctuation">(</span>im<span class="token punctuation">,</span> top<span class="token punctuation">,</span> bottom<span class="token punctuation">,</span> left<span class="token punctuation">,</span> right<span class="token punctuation">,</span> cv2<span class="token punctuation">.</span>BORDER_CONSTANT<span class="token punctuation">,</span> value<span class="token operator">=</span>pad_color<span class="token punctuation">)</span>  <span class="token comment"># add border</span>

<span class="token keyword">if</span> info_need <span class="token keyword">is</span> <span class="token boolean">True</span><span class="token punctuation">:</span>
    <span class="token keyword">return</span> im<span class="token punctuation">,</span> ratio<span class="token punctuation">,</span> <span class="token punctuation">(</span>dw<span class="token punctuation">,</span> dh<span class="token punctuation">)</span>
<span class="token keyword">else</span><span class="token punctuation">:</span>
    <span class="token keyword">return</span> im

def filter_boxes(boxes, box_confidences, box_class_probs):
“”“Filter boxes with object threshold.
“””
box_confidences = box_confidences.reshape(-1)
candidate, class_num = box_class_probs.shape

class_max_score <span class="token operator">=</span> np<span class="token punctuation">.</span><span class="token builtin">max</span><span class="token punctuation">(</span>box_class_probs<span class="token punctuation">,</span> axis<span class="token operator">=</span><span class="token operator">-</span><span class="token number">1</span><span class="token punctuation">)</span>
classes <span class="token operator">=</span> np<span class="token punctuation">.</span>argmax<span class="token punctuation">(</span>box_class_probs<span class="token punctuation">,</span> axis<span class="token operator">=</span><span class="token operator">-</span><span class="token number">1</span><span class="token punctuation">)</span>

_class_pos <span class="token operator">=</span> np<span class="token punctuation">.</span>where<span class="token punctuation">(</span>class_max_score <span class="token operator">*</span> box_confidences <span class="token operator">&gt;=</span> OBJ_THRESH<span class="token punctuation">)</span>
scores <span class="token operator">=</span> <span class="token punctuation">(</span>class_max_score <span class="token operator">*</span> box_confidences<span class="token punctuation">)</span><span class="token punctuation">[</span>_class_pos<span class="token punctuation">]</span>

boxes <span class="token operator">=</span> boxes<span class="token punctuation">[</span>_class_pos<span class="token punctuation">]</span>
classes <span class="token operator">=</span> classes<span class="token punctuation">[</span>_class_pos<span class="token punctuation">]</span>

<span class="token keyword">return</span> boxes<span class="token punctuation">,</span> classes<span class="token punctuation">,</span> scores

def nms_boxes(boxes, scores):
“”“Suppress non-maximal boxes.
# Returns
keep: ndarray, index of effective boxes.
“””
x = boxes[:, 0]
y = boxes[:, 1]
w = boxes[:, 2] - boxes[:, 0]
h = boxes[:, 3] - boxes[:, 1]

areas <span class="token operator">=</span> w <span class="token operator">*</span> h
order <span class="token operator">=</span> scores<span class="token punctuation">.</span>argsort<span class="token punctuation">(</span><span class="token punctuation">)</span><span class="token punctuation">[</span><span class="token punctuation">:</span><span class="token punctuation">:</span><span class="token operator">-</span><span class="token number">1</span><span class="token punctuation">]</span>

keep <span class="token operator">=</span> <span class="token punctuation">[</span><span class="token punctuation">]</span>
<span class="token keyword">while</span> order<span class="token punctuation">.</span>size <span class="token operator">&gt;</span> <span class="token number">0</span><span class="token punctuation">:</span>
    i <span class="token operator">=</span> order<span class="token punctuation">[</span><span class="token number">0</span><span class="token punctuation">]</span>
    keep<span class="token punctuation">.</span>append<span class="token punctuation">(</span>i<span class="token punctuation">)</span>

    xx1 <span class="token operator">=</span> np<span class="token punctuation">.</span>maximum<span class="token punctuation">(</span>x<span class="token punctuation">[</span>i<span class="token punctuation">]</span><span class="token punctuation">,</span> x<span class="token punctuation">[</span>order<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">:</span><span class="token punctuation">]</span><span class="token punctuation">]</span><span class="token punctuation">)</span>
    yy1 <span class="token operator">=</span> np<span class="token punctuation">.</span>maximum<span class="token punctuation">(</span>y<span class="token punctuation">[</span>i<span class="token punctuation">]</span><span class="token punctuation">,</span> y<span class="token punctuation">[</span>order<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">:</span><span class="token punctuation">]</span><span class="token punctuation">]</span><span class="token punctuation">)</span>
    xx2 <span class="token operator">=</span> np<span class="token punctuation">.</span>minimum<span class="token punctuation">(</span>x<span class="token punctuation">[</span>i<span class="token punctuation">]</span> <span class="token operator">+</span> w<span class="token punctuation">[</span>i<span class="token punctuation">]</span><span class="token punctuation">,</span> x<span class="token punctuation">[</span>order<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">:</span><span class="token punctuation">]</span><span class="token punctuation">]</span> <span class="token operator">+</span> w<span class="token punctuation">[</span>order<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">:</span><span class="token punctuation">]</span><span class="token punctuation">]</span><span class="token punctuation">)</span>
    yy2 <span class="token operator">=</span> np<span class="token punctuation">.</span>minimum<span class="token punctuation">(</span>y<span class="token punctuation">[</span>i<span class="token punctuation">]</span> <span class="token operator">+</span> h<span class="token punctuation">[</span>i<span class="token punctuation">]</span><span class="token punctuation">,</span> y<span class="token punctuation">[</span>order<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">:</span><span class="token punctuation">]</span><span class="token punctuation">]</span> <span class="token operator">+</span> h<span class="token punctuation">[</span>order<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">:</span><span class="token punctuation">]</span><span class="token punctuation">]</span><span class="token punctuation">)</span>

    w1 <span class="token operator">=</span> np<span class="token punctuation">.</span>maximum<span class="token punctuation">(</span><span class="token number">0.0</span><span class="token punctuation">,</span> xx2 <span class="token operator">-</span> xx1 <span class="token operator">+</span> <span class="token number">0.00001</span><span class="token punctuation">)</span>
    h1 <span class="token operator">=</span> np<span class="token punctuation">.</span>maximum<span class="token punctuation">(</span><span class="token number">0.0</span><span class="token punctuation">,</span> yy2 <span class="token operator">-</span> yy1 <span class="token operator">+</span> <span class="token number">0.00001</span><span class="token punctuation">)</span>
    inter <span class="token operator">=</span> w1 <span class="token operator">*</span> h1

    ovr <span class="token operator">=</span> inter <span class="token operator">/</span> <span class="token punctuation">(</span>areas<span class="token punctuation">[</span>i<span class="token punctuation">]</span> <span class="token operator">+</span> areas<span class="token punctuation">[</span>order<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">:</span><span class="token punctuation">]</span><span class="token punctuation">]</span> <span class="token operator">-</span> inter<span class="token punctuation">)</span>
    inds <span class="token operator">=</span> np<span class="token punctuation">.</span>where<span class="token punctuation">(</span>ovr <span class="token operator">&lt;=</span> NMS_THRESH<span class="token punctuation">)</span><span class="token punctuation">[</span><span class="token number">0</span><span class="token punctuation">]</span>
    order <span class="token operator">=</span> order<span class="token punctuation">[</span>inds <span class="token operator">+</span> <span class="token number">1</span><span class="token punctuation">]</span>
keep <span class="token operator">=</span> np<span class="token punctuation">.</span>array<span class="token punctuation">(</span>keep<span class="token punctuation">)</span>
<span class="token keyword">return</span> keep

def softmax(x, axis=None):
x = x - x.max(axis=axis, keepdims=True)
y = np.exp(x)
return y / y.sum(axis=axis, keepdims=True)

def dfl(position):
# Distribution Focal Loss (DFL)
n, c, h, w = position.shape
p_num = 4
mc = c // p_num
y = position.reshape(n, p_num, mc, h, w)
y = softmax(y, 2)
acc_metrix = np.array(range(mc), dtype=float).reshape(1, 1, mc, 1, 1)
y = (y * acc_metrix).sum(2)
return y

def box_process(position):
grid_h, grid_w = position.shape[2:4]
col, row = np.meshgrid(np.arange(0, grid_w), np.arange(0, grid_h))
col = col.reshape(1, 1, grid_h, grid_w)
row = row.reshape(1, 1, grid_h, grid_w)
grid = np.concatenate((col, row), axis=1)
stride = np.array([MODEL_SIZE[1] // grid_h, MODEL_SIZE[0] // grid_w]).reshape(1, 2, 1, 1)

position <span class="token operator">=</span> dfl<span class="token punctuation">(</span>position<span class="token punctuation">)</span>
box_xy <span class="token operator">=</span> grid <span class="token operator">+</span> <span class="token number">0.5</span> <span class="token operator">-</span> position<span class="token punctuation">[</span><span class="token punctuation">:</span><span class="token punctuation">,</span> <span class="token number">0</span><span class="token punctuation">:</span><span class="token number">2</span><span class="token punctuation">,</span> <span class="token punctuation">:</span><span class="token punctuation">,</span> <span class="token punctuation">:</span><span class="token punctuation">]</span>
box_xy2 <span class="token operator">=</span> grid <span class="token operator">+</span> <span class="token number">0.5</span> <span class="token operator">+</span> position<span class="token punctuation">[</span><span class="token punctuation">:</span><span class="token punctuation">,</span> <span class="token number">2</span><span class="token punctuation">:</span><span class="token number">4</span><span class="token punctuation">,</span> <span class="token punctuation">:</span><span class="token punctuation">,</span> <span class="token punctuation">:</span><span class="token punctuation">]</span>
xyxy <span class="token operator">=</span> np<span class="token punctuation">.</span>concatenate<span class="token punctuation">(</span><span class="token punctuation">(</span>box_xy <span class="token operator">*</span> stride<span class="token punctuation">,</span> box_xy2 <span class="token operator">*</span> stride<span class="token punctuation">)</span><span class="token punctuation">,</span> axis<span class="token operator">=</span><span class="token number">1</span><span class="token punctuation">)</span>

<span class="token keyword">return</span> xyxy

def post_process(input_data):
boxes, scores, classes_conf = [], [], []
defualt_branch = 3
pair_per_branch = len(input_data) // defualt_branch
# Python 忽略 score_sum 输出
for i in range(defualt_branch):
boxes.append(box_process(input_data[pair_per_branch i]))
classes_conf.append(input_data[pair_per_branch i + 1])
scores.append(np.ones_like(input_data[pair_per_branch * i + 1][:, :1, :, :], dtype=np.float32))

<span class="token keyword">def</span> <span class="token function">sp_flatten</span><span class="token punctuation">(</span>_in<span class="token punctuation">)</span><span class="token punctuation">:</span>
    ch <span class="token operator">=</span> _in<span class="token punctuation">.</span>shape<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">]</span>
    _in <span class="token operator">=</span> _in<span class="token punctuation">.</span>transpose<span class="token punctuation">(</span><span class="token number">0</span><span class="token punctuation">,</span> <span class="token number">2</span><span class="token punctuation">,</span> <span class="token number">3</span><span class="token punctuation">,</span> <span class="token number">1</span><span class="token punctuation">)</span>
    <span class="token keyword">return</span> _in<span class="token punctuation">.</span>reshape<span class="token punctuation">(</span><span class="token operator">-</span><span class="token number">1</span><span class="token punctuation">,</span> ch<span class="token punctuation">)</span>

boxes <span class="token operator">=</span> <span class="token punctuation">[</span>sp_flatten<span class="token punctuation">(</span>_v<span class="token punctuation">)</span> <span class="token keyword">for</span> _v <span class="token keyword">in</span> boxes<span class="token punctuation">]</span>
classes_conf <span class="token operator">=</span> <span class="token punctuation">[</span>sp_flatten<span class="token punctuation">(</span>_v<span class="token punctuation">)</span> <span class="token keyword">for</span> _v <span class="token keyword">in</span> classes_conf<span class="token punctuation">]</span>
scores <span class="token operator">=</span> <span class="token punctuation">[</span>sp_flatten<span class="token punctuation">(</span>_v<span class="token punctuation">)</span> <span class="token keyword">for</span> _v <span class="token keyword">in</span> scores<span class="token punctuation">]</span>

boxes <span class="token operator">=</span> np<span class="token punctuation">.</span>concatenate<span class="token punctuation">(</span>boxes<span class="token punctuation">)</span>
classes_conf <span class="token operator">=</span> np<span class="token punctuation">.</span>concatenate<span class="token punctuation">(</span>classes_conf<span class="token punctuation">)</span>
scores <span class="token operator">=</span> np<span class="token punctuation">.</span>concatenate<span class="token punctuation">(</span>scores<span class="token punctuation">)</span>

<span class="token comment"># filter according to threshold</span>
boxes<span class="token punctuation">,</span> classes<span class="token punctuation">,</span> scores <span class="token operator">=</span> filter_boxes<span class="token punctuation">(</span>boxes<span class="token punctuation">,</span> scores<span class="token punctuation">,</span> classes_conf<span class="token punctuation">)</span>

<span class="token comment"># nms</span>
nboxes<span class="token punctuation">,</span> nclasses<span class="token punctuation">,</span> nscores <span class="token operator">=</span> <span class="token punctuation">[</span><span class="token punctuation">]</span><span class="token punctuation">,</span> <span class="token punctuation">[</span><span class="token punctuation">]</span><span class="token punctuation">,</span> <span class="token punctuation">[</span><span class="token punctuation">]</span>
<span class="token keyword">for</span> c <span class="token keyword">in</span> <span class="token builtin">set</span><span class="token punctuation">(</span>classes<span class="token punctuation">)</span><span class="token punctuation">:</span>
    inds <span class="token operator">=</span> np<span class="token punctuation">.</span>where<span class="token punctuation">(</span>classes <span class="token operator">==</span> c<span class="token punctuation">)</span>
    b <span class="token operator">=</span> boxes<span class="token punctuation">[</span>inds<span class="token punctuation">]</span>
    c <span class="token operator">=</span> classes<span class="token punctuation">[</span>inds<span class="token punctuation">]</span>
    s <span class="token operator">=</span> scores<span class="token punctuation">[</span>inds<span class="token punctuation">]</span>
    keep <span class="token operator">=</span> nms_boxes<span class="token punctuation">(</span>b<span class="token punctuation">,</span> s<span class="token punctuation">)</span>

    <span class="token keyword">if</span> <span class="token builtin">len</span><span class="token punctuation">(</span>keep<span class="token punctuation">)</span> <span class="token operator">!=</span> <span class="token number">0</span><span class="token punctuation">:</span>
        nboxes<span class="token punctuation">.</span>append<span class="token punctuation">(</span>b<span class="token punctuation">[</span>keep<span class="token punctuation">]</span><span class="token punctuation">)</span>
        nclasses<span class="token punctuation">.</span>append<span class="token punctuation">(</span>c<span class="token punctuation">[</span>keep<span class="token punctuation">]</span><span class="token punctuation">)</span>
        nscores<span class="token punctuation">.</span>append<span class="token punctuation">(</span>s<span class="token punctuation">[</span>keep<span class="token punctuation">]</span><span class="token punctuation">)</span>

<span class="token keyword">if</span> <span class="token keyword">not</span> nclasses <span class="token keyword">and</span> <span class="token keyword">not</span> nscores<span class="token punctuation">:</span>
    <span class="token keyword">return</span> <span class="token boolean">None</span><span class="token punctuation">,</span> <span class="token boolean">None</span><span class="token punctuation">,</span> <span class="token boolean">None</span>

boxes <span class="token operator">=</span> np<span class="token punctuation">.</span>concatenate<span class="token punctuation">(</span>nboxes<span class="token punctuation">)</span>
classes <span class="token operator">=</span> np<span class="token punctuation">.</span>concatenate<span class="token punctuation">(</span>nclasses<span class="token punctuation">)</span>
scores <span class="token operator">=</span> np<span class="token punctuation">.</span>concatenate<span class="token punctuation">(</span>nscores<span class="token punctuation">)</span>

<span class="token keyword">return</span> boxes<span class="token punctuation">,</span> classes<span class="token punctuation">,</span> scores

def draw_detections(img, left, top, right, bottom, score, class_id):
“”"
Draws bounding boxes and labels on the input image based on the detected objects.
Args:
img: The input image to draw detections on.
box: Detected bounding box.
score: Corresponding detection score.
class_id: Class ID for the detected object.
Returns:
None
“”"

<span class="token comment"># Retrieve the color for the class ID</span>
color <span class="token operator">=</span> color_palette<span class="token punctuation">[</span>class_id<span class="token punctuation">]</span>

<span class="token comment"># Draw the bounding box on the image</span>
cv2<span class="token punctuation">.</span>rectangle<span class="token punctuation">(</span>img<span class="token punctuation">,</span> <span class="token punctuation">(</span><span class="token builtin">int</span><span class="token punctuation">(</span>left<span class="token punctuation">)</span><span class="token punctuation">,</span> <span class="token builtin">int</span><span class="token punctuation">(</span>top<span class="token punctuation">)</span><span class="token punctuation">)</span><span class="token punctuation">,</span> <span class="token punctuation">(</span><span class="token builtin">int</span><span class="token punctuation">(</span>right<span class="token punctuation">)</span><span class="token punctuation">,</span> <span class="token builtin">int</span><span class="token punctuation">(</span>bottom<span class="token punctuation">)</span><span class="token punctuation">)</span><span class="token punctuation">,</span> color<span class="token punctuation">,</span> <span class="token number">2</span><span class="token punctuation">)</span>

<span class="token comment"># Create the label text with class name and score</span>
label <span class="token operator">=</span> <span class="token string-interpolation"><span class="token string">f"</span><span class="token interpolation"><span class="token punctuation">{<!-- --></span>CLASSES<span class="token punctuation">[</span>class_id<span class="token punctuation">]</span><span class="token punctuation">}</span></span><span class="token string">: </span><span class="token interpolation"><span class="token punctuation">{<!-- --></span>score<span class="token punctuation">:</span><span class="token format-spec">.2f</span><span class="token punctuation">}</span></span><span class="token string">"</span></span>

<span class="token comment"># Calculate the dimensions of the label text</span>
<span class="token punctuation">(</span>label_width<span class="token punctuation">,</span> label_height<span class="token punctuation">)</span><span class="token punctuation">,</span> _ <span class="token operator">=</span> cv2<span class="token punctuation">.</span>getTextSize<span class="token punctuation">(</span>label<span class="token punctuation">,</span> cv2<span class="token punctuation">.</span>FONT_HERSHEY_SIMPLEX<span class="token punctuation">,</span> <span class="token number">0.5</span><span class="token punctuation">,</span> <span class="token number">1</span><span class="token punctuation">)</span>

<span class="token comment"># Calculate the position of the label text</span>
label_x <span class="token operator">=</span> left
label_y <span class="token operator">=</span> top <span class="token operator">-</span> <span class="token number">10</span> <span class="token keyword">if</span> top <span class="token operator">-</span> <span class="token number">10</span> <span class="token operator">&gt;</span> label_height <span class="token keyword">else</span> top <span class="token operator">+</span> <span class="token number">10</span>

<span class="token comment"># Draw a filled rectangle as the background for the label text</span>
cv2<span class="token punctuation">.</span>rectangle<span class="token punctuation">(</span>img<span class="token punctuation">,</span> <span class="token punctuation">(</span>label_x<span class="token punctuation">,</span> label_y <span class="token operator">-</span> label_height<span class="token punctuation">)</span><span class="token punctuation">,</span> <span class="token punctuation">(</span>label_x <span class="token operator">+</span> label_width<span class="token punctuation">,</span> label_y <span class="token operator">+</span> label_height<span class="token punctuation">)</span><span class="token punctuation">,</span> color<span class="token punctuation">,</span>
              cv2<span class="token punctuation">.</span>FILLED<span class="token punctuation">)</span>

<span class="token comment"># Draw the label text on the image</span>
cv2<span class="token punctuation">.</span>putText<span class="token punctuation">(</span>img<span class="token punctuation">,</span> label<span class="token punctuation">,</span> <span class="token punctuation">(</span>label_x<span class="token punctuation">,</span> label_y<span class="token punctuation">)</span><span class="token punctuation">,</span> cv2<span class="token punctuation">.</span>FONT_HERSHEY_SIMPLEX<span class="token punctuation">,</span> <span class="token number">0.5</span><span class="token punctuation">,</span> <span class="token punctuation">(</span><span class="token number">0</span><span class="token punctuation">,</span> <span class="token number">0</span><span class="token punctuation">,</span> <span class="token number">0</span><span class="token punctuation">)</span><span class="token punctuation">,</span> <span class="token number">1</span><span class="token punctuation">,</span> cv2<span class="token punctuation">.</span>LINE_AA<span class="token punctuation">)</span>

def draw(image, boxes, scores, classes):
img_h, img_w = image.shape[:2]
# Calculate scaling factors for bounding box coordinates
x_factor = img_w / MODEL_SIZE[0]
y_factor = img_h / MODEL_SIZE[1]

<span class="token keyword">for</span> box<span class="token punctuation">,</span> score<span class="token punctuation">,</span> cl <span class="token keyword">in</span> <span class="token builtin">zip</span><span class="token punctuation">(</span>boxes<span class="token punctuation">,</span> scores<span class="token punctuation">,</span> classes<span class="token punctuation">)</span><span class="token punctuation">:</span>
    x1<span class="token punctuation">,</span> y1<span class="token punctuation">,</span> x2<span class="token punctuation">,</span> y2 <span class="token operator">=</span> <span class="token punctuation">[</span><span class="token builtin">int</span><span class="token punctuation">(</span>_b<span class="token punctuation">)</span> <span class="token keyword">for</span> _b <span class="token keyword">in</span> box<span class="token punctuation">]</span>

    left <span class="token operator">=</span> <span class="token builtin">int</span><span class="token punctuation">(</span>x1 <span class="token operator">*</span> x_factor<span class="token punctuation">)</span>
    top <span class="token operator">=</span> <span class="token builtin">int</span><span class="token punctuation">(</span>y1 <span class="token operator">*</span> y_factor<span class="token punctuation">)</span>
    right <span class="token operator">=</span> <span class="token builtin">int</span><span class="token punctuation">(</span>x2 <span class="token operator">*</span> x_factor<span class="token punctuation">)</span>
    bottom <span class="token operator">=</span> <span class="token builtin">int</span><span class="token punctuation">(</span>y2 <span class="token operator">*</span> y_factor<span class="token punctuation">)</span>

    <span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'class: {}, score: {}'</span><span class="token punctuation">.</span><span class="token builtin">format</span><span class="token punctuation">(</span>CLASSES<span class="token punctuation">[</span>cl<span class="token punctuation">]</span><span class="token punctuation">,</span> score<span class="token punctuation">)</span><span class="token punctuation">)</span>
    <span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'box coordinate left,top,right,down: [{}, {}, {}, {}]'</span><span class="token punctuation">.</span><span class="token builtin">format</span><span class="token punctuation">(</span>left<span class="token punctuation">,</span> top<span class="token punctuation">,</span> right<span class="token punctuation">,</span> bottom<span class="token punctuation">)</span><span class="token punctuation">)</span>

    <span class="token comment"># Retrieve the color for the class ID</span>

    draw_detections<span class="token punctuation">(</span>image<span class="token punctuation">,</span> left<span class="token punctuation">,</span> top<span class="token punctuation">,</span> right<span class="token punctuation">,</span> bottom<span class="token punctuation">,</span> score<span class="token punctuation">,</span> cl<span class="token punctuation">)</span>

    <span class="token comment"># cv2.rectangle(image, (left, top), (right, bottom), color, 2)</span>
    <span class="token comment"># cv2.putText(image, '{0} {1:.2f}'.format(CLASSES[cl], score),</span>
    <span class="token comment">#             (left, top - 6),</span>
    <span class="token comment">#             cv2.FONT_HERSHEY_SIMPLEX,</span>
    <span class="token comment">#             0.6, (0, 0, 255), 2)</span>

if name == ‘main’:

<span class="token comment"># 创建RKNN对象</span>
rknn_lite <span class="token operator">=</span> RKNNLite<span class="token punctuation">(</span><span class="token punctuation">)</span>

<span class="token comment"># 加载RKNN模型</span>
<span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'--&gt; Load RKNN model'</span><span class="token punctuation">)</span>
ret <span class="token operator">=</span> rknn_lite<span class="token punctuation">.</span>load_rknn<span class="token punctuation">(</span>RKNN_MODEL<span class="token punctuation">)</span>
<span class="token keyword">if</span> ret <span class="token operator">!=</span> <span class="token number">0</span><span class="token punctuation">:</span>
    <span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'Load RKNN model failed'</span><span class="token punctuation">)</span>
    exit<span class="token punctuation">(</span>ret<span class="token punctuation">)</span>
<span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'done'</span><span class="token punctuation">)</span>

<span class="token comment"># 初始化 runtime 环境</span>
<span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'--&gt; Init runtime environment'</span><span class="token punctuation">)</span>
<span class="token comment"># run on RK356x/RK3588 with Debian OS, do not need specify target.</span>
ret <span class="token operator">=</span> rknn_lite<span class="token punctuation">.</span>init_runtime<span class="token punctuation">(</span><span class="token punctuation">)</span>
<span class="token keyword">if</span> ret <span class="token operator">!=</span> <span class="token number">0</span><span class="token punctuation">:</span>
    <span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'Init runtime environment failed!'</span><span class="token punctuation">)</span>
    exit<span class="token punctuation">(</span>ret<span class="token punctuation">)</span>
<span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'done'</span><span class="token punctuation">)</span>

<span class="token comment"># 数据处理</span>
img_list <span class="token operator">=</span> os<span class="token punctuation">.</span>listdir<span class="token punctuation">(</span>IMG_FOLDER<span class="token punctuation">)</span>
<span class="token keyword">for</span> i <span class="token keyword">in</span> <span class="token builtin">range</span><span class="token punctuation">(</span><span class="token builtin">len</span><span class="token punctuation">(</span>img_list<span class="token punctuation">)</span><span class="token punctuation">)</span><span class="token punctuation">:</span>
    img_name <span class="token operator">=</span> img_list<span class="token punctuation">[</span>i<span class="token punctuation">]</span>
    img_path <span class="token operator">=</span> os<span class="token punctuation">.</span>path<span class="token punctuation">.</span>join<span class="token punctuation">(</span>IMG_FOLDER<span class="token punctuation">,</span> img_name<span class="token punctuation">)</span>
    <span class="token keyword">if</span> <span class="token keyword">not</span> os<span class="token punctuation">.</span>path<span class="token punctuation">.</span>exists<span class="token punctuation">(</span>img_path<span class="token punctuation">)</span><span class="token punctuation">:</span>
        <span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">"{} is not found"</span><span class="token punctuation">,</span> img_name<span class="token punctuation">)</span>
        <span class="token keyword">continue</span>
    img_src <span class="token operator">=</span> cv2<span class="token punctuation">.</span>imread<span class="token punctuation">(</span>img_path<span class="token punctuation">)</span>
    <span class="token keyword">if</span> img_src <span class="token keyword">is</span> <span class="token boolean">None</span><span class="token punctuation">:</span>
        <span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">"文件不存在\n"</span><span class="token punctuation">)</span>

    <span class="token comment"># Due to rga init with (0,0,0), we using pad_color (0,0,0) instead of (114, 114, 114)</span>
    pad_color <span class="token operator">=</span> <span class="token punctuation">(</span><span class="token number">0</span><span class="token punctuation">,</span> <span class="token number">0</span><span class="token punctuation">,</span> <span class="token number">0</span><span class="token punctuation">)</span>
    img <span class="token operator">=</span> letter_box<span class="token punctuation">(</span>im<span class="token operator">=</span>img_src<span class="token punctuation">.</span>copy<span class="token punctuation">(</span><span class="token punctuation">)</span><span class="token punctuation">,</span> new_shape<span class="token operator">=</span><span class="token punctuation">(</span>MODEL_SIZE<span class="token punctuation">[</span><span class="token number">1</span><span class="token punctuation">]</span><span class="token punctuation">,</span> MODEL_SIZE<span class="token punctuation">[</span><span class="token number">0</span><span class="token punctuation">]</span><span class="token punctuation">)</span><span class="token punctuation">,</span> pad_color<span class="token operator">=</span><span class="token punctuation">(</span><span class="token number">0</span><span class="token punctuation">,</span> <span class="token number">0</span><span class="token punctuation">,</span> <span class="token number">0</span><span class="token punctuation">)</span><span class="token punctuation">)</span>
    <span class="token comment"># img = cv2.resize(img_src, (640, 512), interpolation=cv2.INTER_LINEAR) # direct resize</span>
    <span class="token builtin">input</span> <span class="token operator">=</span> np<span class="token punctuation">.</span>expand_dims<span class="token punctuation">(</span>img<span class="token punctuation">,</span> axis<span class="token operator">=</span><span class="token number">0</span><span class="token punctuation">)</span>

    outputs <span class="token operator">=</span> rknn_lite<span class="token punctuation">.</span>inference<span class="token punctuation">(</span><span class="token punctuation">[</span><span class="token builtin">input</span><span class="token punctuation">]</span><span class="token punctuation">)</span>

    boxes<span class="token punctuation">,</span> classes<span class="token punctuation">,</span> scores <span class="token operator">=</span> post_process<span class="token punctuation">(</span>outputs<span class="token punctuation">)</span>

    img_p <span class="token operator">=</span> img_src<span class="token punctuation">.</span>copy<span class="token punctuation">(</span><span class="token punctuation">)</span>

    <span class="token keyword">if</span> boxes <span class="token keyword">is</span> <span class="token keyword">not</span> <span class="token boolean">None</span><span class="token punctuation">:</span>
        draw<span class="token punctuation">(</span>img_p<span class="token punctuation">,</span> boxes<span class="token punctuation">,</span> scores<span class="token punctuation">,</span> classes<span class="token punctuation">)</span>

    <span class="token comment"># 保存结果</span>
    <span class="token keyword">if</span> <span class="token keyword">not</span> os<span class="token punctuation">.</span>path<span class="token punctuation">.</span>exists<span class="token punctuation">(</span>RESULT_PATH<span class="token punctuation">)</span><span class="token punctuation">:</span>
        os<span class="token punctuation">.</span>mkdir<span class="token punctuation">(</span>RESULT_PATH<span class="token punctuation">)</span>

    result_path <span class="token operator">=</span> os<span class="token punctuation">.</span>path<span class="token punctuation">.</span>join<span class="token punctuation">(</span>RESULT_PATH<span class="token punctuation">,</span> img_name<span class="token punctuation">)</span>
    cv2<span class="token punctuation">.</span>imwrite<span class="token punctuation">(</span>result_path<span class="token punctuation">,</span> img_p<span class="token punctuation">)</span>
    <span class="token keyword">print</span><span class="token punctuation">(</span><span class="token string">'Detection result save to {}'</span><span class="token punctuation">.</span><span class="token builtin">format</span><span class="token punctuation">(</span>result_path<span class="token punctuation">)</span><span class="token punctuation">)</span>

    <span class="token keyword">pass</span>

<span class="token comment"># cv2.imshow("full post process result", img_p)</span>

rknn_lite<span class="token punctuation">.</span>release<span class="token punctuation">(</span><span class="token punctuation">)</span>
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15
  • 16
  • 17
  • 18
  • 19
  • 20
  • 21
  • 22
  • 23
  • 24
  • 25
  • 26
  • 27
  • 28
  • 29
  • 30
  • 31
  • 32
  • 33
  • 34
  • 35
  • 36
  • 37
  • 38
  • 39
  • 40
  • 41
  • 42
  • 43
  • 44
  • 45
  • 46
  • 47
  • 48
  • 49
  • 50
  • 51
  • 52
  • 53
  • 54
  • 55
  • 56
  • 57
  • 58
  • 59
  • 60
  • 61
  • 62
  • 63
  • 64
  • 65
  • 66
  • 67
  • 68
  • 69
  • 70
  • 71
  • 72
  • 73
  • 74
  • 75
  • 76
  • 77
  • 78
  • 79
  • 80
  • 81
  • 82
  • 83
  • 84
  • 85
  • 86
  • 87
  • 88
  • 89
  • 90
  • 91
  • 92
  • 93
  • 94
  • 95
  • 96
  • 97
  • 98
  • 99
  • 100
  • 101
  • 102
  • 103
  • 104
  • 105
  • 106
  • 107
  • 108
  • 109
  • 110
  • 111
  • 112
  • 113
  • 114
  • 115
  • 116
  • 117
  • 118
  • 119
  • 120
  • 121
  • 122
  • 123
  • 124
  • 125
  • 126
  • 127
  • 128
  • 129
  • 130
  • 131
  • 132
  • 133
  • 134
  • 135
  • 136
  • 137
  • 138
  • 139
  • 140
  • 141
  • 142
  • 143
  • 144
  • 145
  • 146
  • 147
  • 148
  • 149
  • 150
  • 151
  • 152
  • 153
  • 154
  • 155
  • 156
  • 157
  • 158
  • 159
  • 160
  • 161
  • 162
  • 163
  • 164
  • 165
  • 166
  • 167
  • 168
  • 169
  • 170
  • 171
  • 172
  • 173
  • 174
  • 175
  • 176
  • 177
  • 178
  • 179
  • 180
  • 181
  • 182
  • 183
  • 184
  • 185
  • 186
  • 187
  • 188
  • 189
  • 190
  • 191
  • 192
  • 193
  • 194
  • 195
  • 196
  • 197
  • 198
  • 199
  • 200
  • 201
  • 202
  • 203
  • 204
  • 205
  • 206
  • 207
  • 208
  • 209
  • 210
  • 211
  • 212
  • 213
  • 214
  • 215
  • 216
  • 217
  • 218
  • 219
  • 220
  • 221
  • 222
  • 223
  • 224
  • 225
  • 226
  • 227
  • 228
  • 229
  • 230
  • 231
  • 232
  • 233
  • 234
  • 235
  • 236
  • 237
  • 238
  • 239
  • 240
  • 241
  • 242
  • 243
  • 244
  • 245
  • 246
  • 247
  • 248
  • 249
  • 250
  • 251
  • 252
  • 253
  • 254
  • 255
  • 256
  • 257
  • 258
  • 259
  • 260
  • 261
  • 262
  • 263
  • 264
  • 265
  • 266
  • 267
  • 268
  • 269
  • 270
  • 271
  • 272
  • 273
  • 274
  • 275
  • 276
  • 277
  • 278
  • 279
  • 280
  • 281
  • 282
  • 283
  • 284
  • 285
  • 286
  • 287
  • 288
  • 289
  • 290
  • 291
  • 292
  • 293
  • 294
  • 295
  • 296
  • 297
  • 298
  • 299
  • 300
  • 301
  • 302
  • 303
  • 304
  • 305
  • 306
  • 307
  • 308
  • 309
  • 310
  • 311
  • 312
  • 313
  • 314

使用下面命令来执行

python rk3588_test-2.py

 
 

 
 
  • 1

会输出检测的结果如下图所示

!外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传

yolov8 rknn3588混合量化
09-26
yolov8 rknn3588混合量化
rk3588部署yolov8视频目标检测教程
jngwe的博客
04-30 6894
解压opencv和opencv-contrib 代码，在板卡上新建一个opencv4_install文件夹，用于安装opencv，再进入解压后的opencv4.4.0文件夹， 新建文件夹 build，进入build文件夹，打开终端执行以下命令。patch文件我已经放到百度云opencv4.4/patch_文件夹下，可以下载放到对应位置后，在重新执行上述指令进行编译，直到无报错为止，再执行make指令进行编译。执行完会在终端显示 done!编译的时候，会报错误，缺一些依赖库，参照下面的链接进行解决。
RK3588部署YOLOv8
m0_48979117的博客
01-16 1万+
本人基于YOLOv8改进水了一篇论文，为了增加工作量在RK3588上部署了改进的算法。根据网上的方法一直没有部署成功，最后在参考官方和网络上的方法成功实现算法的部署，因此写下这篇博客进行记录，以防后面忘了怎么部署，也可供网友参考.
RK3588/RK3588s运行yolov8推理速度达到27ms
cg1135217680的博客
09-20 2864
小伙伴们~~我最近做了一个比较有意思的东西，想起来也好久没有写博客了，就记录一下吧。希望和大家一起学习，一起进步！ 我简单介绍一下我最近做的这个东西的经过哈~上个月在B站上看到了一个博主发了一条视频关于边缘计算相关的，他仅仅用了一块巴掌不到的开发板完成了yolo的目标检测功能，而且延时还比较低，成本也仅有100元不到！倍感神奇哇！然后我就去学习了一下，关于NPU计算相关的知识，刚好有小伙伴手里有一块类似的开发板，于是就有了这篇博客。
【YOLOv10部署RK3588】模型训练→转换RKNN→开发板部署
A_l_b_ert的博客
11-11 2729
最近清华发布了不需要NMS的YOLO检测算法：YOLOv10，感觉挺新奇，准备尝试一下，并部署到国产RK3588开发板测试下效果。此前博主已经发布过YOLOv8n的相关内容，，其实具体实现方法都是差不多的，所以内容直接从YOLOv8中copy过来，具体不同细节再详细论述。
YOLOv8部署在RK3588上（亲测可行）
qq_42541521的博客
06-11 1427
yolov8模型部署在rk3588设备上，模型转化，环境搭建等
yolov8部署到RK3588_并运行示例12.4_笔记2
m0_60657960的博客
12-04 1947
将下面的so文件复制到/usr/lib/下。
【YOLOv8部署至RK3588】模型训练→转换RKNN→开发板部署
热门推荐
A_l_b_ert的博客
08-27 1万+
之前博主发布过YOLOv8转RKNN模型并在开发板上部署的流程，但是遇到了许多问题，发现很难修补，遂决定在官方项目下进行模型转换与部署（之前的转换是在Github个人博主的项目下进行的）OK，进入正题，模型转换需要以下工具：ultralytics_yolov8、rknn_model_zoo、rknn-toolkit2，现在详细介绍这三者的关系。以上三个文件均与模型训练无关，而是用于模型转换的，其中，ultralytics_yolov8文件不是必需的，但最好还是下载一份，原因后面会讲；。
YOLOv8部署在RKNN3588上
qq_43571475的博客
05-31 9825
并且我将bbox分支的reshape后面的算子全部去除，因为在板子上运行的时候老是报错，由于这个地方在源码中就是将bbox的预测输出变成xyhw的输出格式，于是干脆砍掉，在后处理的过程中修改即可。在转为rknn模型的过程中，如果打开了量化，类别得分会全部被量化为0，所以单输出的onnx转成rknn会有问题（如果有大佬有解决办法，并且成功对单输出的模型进行后处理的，欢迎讨论分享）注意，onnx要转成rknn模型，目前不支持太高版本的opset算子库，YOLOv8默认是opset=17，我们需要修改成12。
端侧AI加速：YOLOv11在RK3588芯片的INT8量化部署全解析.pdf
02-17
YOLOv11结合RK3588芯片的INT8量化部署，不仅能够大幅提升目标检测的速度，而且能在保证精度的前提下，满足端侧设备实时检测的需求。这一技术的深入解析，对目标检测技术的研究和应用具有重要的指导意义。
YOLOV8模型转换-ONNX-RKNN
10-12
YOLOV8模型转换至ONNX-RKNN的流程是一个典型的深度学习模型部署步骤，尤其在嵌入式设备如RK3588上运行时，这样的优化和转换至关重要。YOLO（You Only Look Once）系列是实时目标检测系统，而YOLOV8作为其最新版本，...
yolov8目标检测在rk3588部署全过程
最新发布
02-24
### YOLOv8 目标检测 RK3588 部署教程 #### 环境配置 在RK3588平台上部署YOLOv8模型之前，需先完成环境配置工作。这一步骤涉及安装必要的软件包和库文件，确保硬件设备能够支持后续的操作流程[^1]。 #### 修改模型...
yolov5部署在rk3588,pt转onnx，onnx转rknn（附安装包）
qq_42541521的博客
05-21 2725
执行完后会生成目录。
rk3588部署yolov5
qq_59164231的博客
08-29 7208
YOLOv5 训练得到pt文件，pt-->onnx-->rknn进行推理。
RK3588：Rknn模型转换及部署YOLOV8模型教程--飞凌
weixin_47464132的博客
09-25 3381
rk3588部署YOLOV8模型
yolov5训练并生成rknn模型部署在RK3588开发板上，实现NPU加速推理
m0_62919535的博客
03-24 1万+
RK3588是瑞芯微（Rockchip）公司推出的一款高性能、低功耗的集成电路芯片。它采用了先进的28纳米工艺技术，并配备了八核心的ARM Cortex-A76和Cortex-A55处理器，以及ARM Mali-G76 GPU。该芯片支持多种接口和功能，适用于广泛的应用领域。本篇为yolov5部署在RK3588的教程。接着将训练好的best.pt放在工程文件夹下，使用yolov5工程中的export.py将其转换为onnx模型。
深入探索：RK3588平台开发系列——基于YOLOv5的目标检测项目实战
一直在水些技术小文
12-27 1048
YOLOv5是一种高性能的目标检测算法，具有快速、准确的特点。它通过一次前向传播实现对目标的检测，适用于嵌入式系统等有计算资源限制的环境。在RK3588平台上应用YOLOv5，将为项目实战注入强大的视觉智能。通过本文的实例演示，我们深入探索了在RK3588平台上开发目标检测项目的全过程。YOLOv5作为一种轻量级、高性能的目标检测算法，为嵌入式系统提供了强大的视觉智能支持。在未来的技术分享中，我们将继续深入挖掘RK3588平台的开发技巧，为大家呈现更多关于嵌入式系统、深度学习应用等领域的精彩内容。
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值