luolinll1212的专栏

ubuntu

triton

zip 压缩和分卷压缩

1.下载paddlepaddle安装包https://paddleinference.paddlepaddle.org.cn/master/user_guides/download_lib.html#python选择：Jetpack4.4(4.5): nv_jetson-cuda10.2-trt7-all2.安装apt-get install libgeos-dev # shapely 包git clone https://github.com/PaddlePaddle/PaddleOCR.gi

1.安装配置cuda 10.0cudnntensorrtdeepstream 5.1

ffmpeg

1，安装sudo apt-get install git gcc make cmake yasm python p7zip-full libssl-dev pkg-config autoconfautomake build-essential -ysudo apt-get -y --force-yes install libass-devsudo apt-get -y --force-yes install libtheora-dev libtool libva-dev libvdpau-devs

1，介绍2，安装3，使用

1.介绍商汤2.安装pip install mmcv-full mmdet mmocr mmsegmentation mmcv3.demo项目， 数据集CHASE DB1# 1.拉项目git clone https://github.com/open-mmlab/mmsegmentation.gitcd mmsegmentationmkdir datawget https://staffnet.kingston.ac.uk/~ku15565/CHASE_DB1/assets/CHASED

1.ubuntusudo apt install gdb gdbserver2.vscode安装插件扩展 -> Remote Development3.远程开发

1.linaro官网下载交叉编译工具，选择gcc-linaro-10.2.1-2021.02-x86_64_aarch64-linux-gnu.tar.xz，解压即可用

1.介绍2.安装2.1 linuxwget http://www.ffmpeg.org/releases/ffmpeg-3.1.tar.gztar -zxvf ffmpeg-3.1.tar.gzcd ffmpeg-3.1./configure --prefix=/usr/local/ffmpegmake && make install# 添加环境变量vi /etc/profileexport PATH=$PATH:/usr/local/ffmpeg/binsource

0,介绍    NVIDIA/DALI(NVIDIA Data Loading Library) 库包含了用于加速深度学习应用的数据预处理的高度优化的构建模块(highly optimized building blocks)和执行引擎(execution engine).    深度学习应用中，往往需要复杂的、多阶段的数据预处理管道. 这些数据管道主要是在CPU上执行的数据密集型操作(compute-intensive o

input = []for i in range(10): x = np.random.rand(3, 224, 224) input.append(x)t1 = time.time()input = np.asarray(input)input = torch.from_numpy(input)

anaconda

ubuntu18.04 安装cuda

2，编译git clone https://github.com/apache/tvm.gitcd tvmmkdir buildcp cmake/config.cmake buildcd buildcmake ..make -j 8

vulkan

export PYTHONPATH="/path/that/contains/your/so":"${PYTHONPATH}"

1,pybind11简介2,代码

编译darknetgit clone https://github.com/AlexeyAB/darknet.gitcd darknetvim makefile LIBSO=0 -> LIBSO=1make # 生成libdarknet.so

1，准备需要下载三个包，分别是glm,glfw,vulkanglm，地址glfw，地址vulkan，地址以上为已经编译。2，VS2015创建项目文件 -> 新建 -> 控制台应用程序。选择，Release x642.1 添加目录glm/includeglfw/includevulkan/include2.2 添加库gflw/libvulkan/lib2.3 添加连接器glfw3.libvulkan-1.lib2.4 添加代码#define GLFW_I

1，介绍Eigen下载地址:https://gitlab.com/libeigen/eigen/-/releases2，安装2.1 编译wget https://gitlab.com/libeigen/eigen/-/archive/3.3.7/eigen-3.3.7.tar.gztar -xf eigen-3.3.7.tar.gzcd eigenmkdir buildcmake ..make -j 8sudo make install2.2 复制目录sudo cp -r /usr

__

1,安装

1，安装

1，介绍2，安装sudo apt -y install p7zip-full bc curl klibc-utils iproute2 genisoimage dosfstools sudo apt -y install squashfs-tools rsync unzip wget findutils xorriso下载，l_openvino_toolkit_p_2020.3.194.tgz，序列号3，应用

1，介绍2，安装2,1 ubuntu 18.04sudo apt -y install p7zip-full bc curl klibc-utils iproute2 genisoimage dosfstools sudo apt -y install squashfs-tools rsync unzip wget findutils xorriso2.2 win10

1，介绍2，安装

1，介绍2，安装

1,介绍    ncnn 是一个为手机端极致优化的高性能神经网络前向计算框架。2,编译git clone https://github.com/Tencent/ncnn.gitcd ncnnmkdir buildcd buildcmake ..make -j 4

1，介绍    MNN是一个轻量级的深度神经网络推理引擎，在端侧加载深度神经网络模型进行推理预测。2，编译git clone https://github.com/alibaba/MNN.gitcd MNNmkdir buildcd buildcmake .. -DMNN_BUILD_CONVERTER=truemake./build_tool.sh...

1，介绍onnx-tensorrt是一个模型推理框架。2，安装sudo apt-get install libprotobuf-dev protobuf-compilersudo apt-get install swiggit clone https://github.com/onnx/onnx-tensorrt.gitgit submodule syncgit submodule update --init --recursive cd onnx-tensorrtmkdir build

onnxruntime

opengl

1，安装sudo apt-get install libprotobuf* protobuf-compiler ninja-buildgit clone https://github.com/NVIDIA-AI-IOT/torch2trtcd torch2trtpython setup.py install --plugins # 本地的虚拟环境报错，#NvInfer.h，找不到def trt_inc_dir(): return "../targets/x86_64-linux-gnu

sudo apt install gdebi-corewget https://download.teamviewer.com/download/linux/teamviewer_amd64.debsudo gdebi teamviewer_amd64.deb

tensorrt

#include <torch/script.h> // One-stop header.#include <iostream>#include <memory>int main(int argc, const char* argv[]) { if (argc != 2) { std::cerr << "usage: example-app <path-to-exported-script-module>\n"; re

1，下载cygwin64http://www.cygwin.com,下载setup-x86_64.exe2，安装镜像源可以选择163（网易的）地址，安装gcc-core、gcc-g++、make、cmake、gdb、binutils，六个包都在Category下的Devel下 。查看版本cygcheck -c cygwin...