qq_35275007的博客

import mathfrom math import fabs, sin, cos, radiansimport matplotlib.pyplot as pltimport numpy as npfrom PIL import Image, ImageDrawimport cv2import randomimport os'''根据列表点，在图中画折线'''def draw_Polyline(img, point_list, color=(0, 0, 255)): ...

import mathfrom math import fabs, sin, cos, radiansimport matplotlib.pyplot as pltimport numpy as npfrom PIL import Image, ImageDrawimport cv2import randomimport os'''根据列表点，在图中画折线'''def draw_Polyline(img, point_list, color=(0, 0, 255)): .

import torchimport torch.nn as nn#__all__ = ['ResNet', 'resnet18', 'resnet34', 'resnet50', 'resnet101',# 'resnet152', 'resnext50_32x4d', 'resnext101_32x8d',# 'wide_resnet50_2', 'wide_resnet101_2']###model_urls = {# 'resne.

SimAM: A Simple, Parameter-Free Attention Module for Convolutional Neural Networks论文链接（已收录于ICML 2021）：http://proceedings.mlr.press/v139/yang21o.htmlcode: https://github.com/ZjjConan/SimAM参考博客：https://blog.youkuaiyun.com/hb_learing/article/details/1190778

这是HRNet的官方代码：https://github.com/HRNet更改它的输出头，就可以变成分类、目标检测、分割、姿态检测等网络我是按照以下代码画的网络结构https://github.com/stefanopini/simple-HRNet/blob/master/models/hrnet.py...

import mathfrom math import fabs, sin, cos, radiansimport matplotlib.pyplot as pltimport numpy as npfrom PIL import Image, ImageDrawimport cv2import randomimport os'''根据列表点，在图中画折线'''def draw_Polyline(img, point_list, color=(0, 0, 255)): .

import mathfrom math import fabs, sin, cos, radiansimport matplotlib.pyplot as pltimport numpy as npfrom PIL import Image, ImageDrawimport cv2import randomimport os#返回一张旋转后的图片def Rotate_img(img, angle): height,width=img.shape[:2] d.

import mathfrom math import fabs, sin, cos, radiansimport matplotlib.pyplot as pltimport numpy as npfrom PIL import Image, ImageDrawimport cv2import randomimport os'''添加一个随机冲击响应'''def add_attack_response(point_list): #制造一个随机数，产生冲击响应 .

import mathimport torchimport torch.nn as nnimport torch.nn.functional as Fimport torch.utils.datafrom torchvision import datasets, transformsfrom torchvision import modelsimport matplotlib.pyplot as plttorch.manual_seed(0)# torch.cuda.manual_se.

Mosaic数据增强什么是Mosaic数据增强方法Yolov4的mosaic数据增强参考了CutMix数据增强方式，理论上具有一定的相似性！CutMix数据增强方式利用两张图片进行拼接。但是mosaic利用了四张图片，根据论文所说其拥有一个巨大的优点是丰富检测物体的背景！且在BN计算的时候一下子会计算四张图片的数据！就像下图这样：实现思路1、每次读取四张图片。2、分别对四张图片进行翻转、缩放、色域变化等，并且按照四个方向位置摆好。3、进行图片的组合和框的组合全部代码全