wulingtian / yolov5_caffe Goto Github PK

yolov5 onnx caffe

License: Other

Makefile 39.41% CMake 1.64% Shell 0.28% Dockerfile 0.05% C++ 48.86% Cuda 3.73% MATLAB 0.53% Python 5.50%

yolov5 onnx caffe

yolov5_caffe's Introduction

环境配置

ubuntu：18.04

cuda：10.0

cudnn：7.6.5

caffe: 1.0

OpenCV：3.4.2

Anaconda3：5.2.0

相关的安装包我已经放到百度云盘，可以从如下链接下载: https://pan.baidu.com/s/17bjiU4H5O36psGrHlFdM7A 密码: br7h

cuda和cudnn的安装

可以参考我的另一篇部署文章（TensorRT int8 量化部署 yolov5s 4.0 模型）

Anaconda安装

chmod +x Anaconda3-5.2.0-Linux-x86_64.sh（从上面百度云盘链接下载）

./Anaconda3-5.2.0-Linux-x86_64.sh

按ENTER，然后按q调至结尾

接受协议　yes

安装路径　使用默认路径

执行安装

在使用的用户.bashrc上添加anaconda路径，比如

export PATH=/home/willer/anaconda3/bin:$PATH

caffe安装

git clone https://github.com/Wulingtian/yolov5_caffe.git

cd yolov5_caffe

命令行输入如下内容：

export CPLUS_INCLUDE_PATH=/home/你的用户名/anaconda3/include/python3.6m

make all -j8

make pycaffe -j8

vim ~/.bashrc

export PYTHONPATH=/home/你的用户名/yolov5_caffe/python:$PYTHONPATH

source ~/.bashrc

编译过程踩过的坑

libstdc++.so.6: version `GLIBCXX_3.4.21' not found

解决方案：搞定 libstdc++.so.6: version `GLIBCXX_3.4.21' not found

ImportError: No module named google.protobuf.internal

解决方案：ImportError: No module named google.protobuf.internal

wrap_python.hpp:50:23: fatal error: pyconfig.h: No such file or dir

解决方案：caffe : /wrap_python.hpp:50:23: fatal error: pyconfig.h: No such file or dir

yolov5s模型转换onnx模型

pip安装onnx和onnx-simplifier

pip install onnx

pip install onnx-simplifier

拉取yolov5官方代码

git clone https://github.com/ultralytics/yolov5.git

训练自己的模型步骤参考yolov5官方介绍，训练完成后我们得到了一个模型文件

cd yolov5

vim models/export.py 修改opset_version为10

python models/export.py --weights 训练得到的模型权重路径 --img-size 训练图片输入尺寸

python -m onnxsim onnx模型名称 yolov5s-simple.onnx 得到最终简化后的onnx模型

onnx模型转换caffe模型

git clone https://github.com/Wulingtian/yolov5_onnx2caffe.git

cd yolov5_onnx2caffe

vim convertCaffe.py

设置onnx_path（上面转换得到的onnx模型），prototxt_path（caffe的prototxt保存路径），caffemodel_path（caffe的caffemodel保存路径）

python convertCaffe.py 得到转换好的caffe模型

caffe模型推理

定位到yolov5_caffe目录下

cd tools

vim caffe_yolov5s.cpp

设置如下参数：

INPUT_W（模型输入宽度）

INPUT_H（模型输入高度）

NUM_CLASS（模型有多少个类别，例如我训练的模型是安全帽检测，只有1类，所以设置为1，不需要加背景类）

NMS_THRESH（做非极大值抑制的阈值）

CONF_THRESH（类别置信度）

prototxt_path（caffe模型的prototxt路径）

caffemodel_path（caffe模型的caffemodel路径）

pic_path（预测图片的路径）

定位到yolov5_caffe目录下

make -j8

cd build

./tools/caffe_yolov5s 输出平均推理时间，以及保存预测图片到当前目录下，至此，部署完成！

yolov5_caffe's People

Contributors

Stargazers

Watchers

yolov5_caffe's Issues

作者您好，请问这个可以转yolov5m的模型吗，如果可以的话需不需要改动哪里

你好,我使用hi3519av100部署yolov5,按照您的步骤走,程序是跑通了,但是最后的输出有问题,能麻烦看一下吗

具体表现为无论我的输入bgr是什么,最后输出都是固定的,而且感觉框都是乱画的

我使用了三张不同的bgr图片,输出都是这样

I don't know which bags are needed. Can you give me one requirements.txt Is a similar document?

It is can not convert onnx to caffe model

hello;
i have build the environment successfully fellow the illustration. when i run python convertCaffe.py; it return as bellow :
Traceback (most recent call last):
File "convertCaffe.py", line 6, in
import caffe
File "/home/bkuser/work2/data/liudongbo/yolov5_caffe/python/caffe/init.py", line 1, in
from .pycaffe import Net, SGDSolver, NesterovSolver, AdaGradSolver, RMSPropSolver, AdaDeltaSolver, AdamSolver, NCCL, Timer
File "/home/bkuser/work2/data/liudongbo/yolov5_caffe/python/caffe/pycaffe.py", line 13, in
from ._caffe import Net, SGDSolver, NesterovSolver, AdaGradSolver,
ImportError: dynamic module does not define module export function (PyInit__caffe)

then i run python2 convertCaffe.py; it return as bellow :
Traceback (most recent call last):
File "convertCaffe.py", line 6, in
import caffe
File "/home/bkuser/work2/data/liudongbo/yolov5_caffe/python/caffe/init.py", line 1, in
from .pycaffe import Net, SGDSolver, NesterovSolver, AdaGradSolver, RMSPropSolver, AdaDeltaSolver, AdamSolver, NCCL, Timer
File "/home/bkuser/work2/data/liudongbo/yolov5_caffe/python/caffe/pycaffe.py", line 13, in
from ._caffe import Net, SGDSolver, NesterovSolver, AdaGradSolver,
ImportError: dynamic module does not define module export function (PyInit__caffe)
bkuser@bk-ai3:~/work2/data/liudongbo/yolov5_onnx2caffe$ python2 convertCaffe.py
Traceback (most recent call last):
File "convertCaffe.py", line 8, in
import onnx
ImportError: No module named onnx

then , i run pip2 install onnx ,but it failed.

could you give me an advice please ?
Yours

the performace

hello，thx for your code，i have test the yolov5s，but the res is not well，could you help me confirm if there is any loss between onnx-caffe？thx again！

The problem about C++11 when make all -j8

when I make all -j8,the problem as:
tools/caffe_yolov5s.cpp:246:5: warning: identifier 'nullptr' is a keyword in C++11 [-Wc++0x-compat]
Net caffe_net(prototxt_path, caffe::TEST, 0, nullptr);
^
tools/caffe_yolov5s.cpp: In function 'std::vector initAnchors()':
tools/caffe_yolov5s.cpp:62:13: error: 'class std::vector' has no member named 'emplace_back'
anchors.emplace_back(anchor);

ubuntu：18.04
cuda：10.0
cudnn：7.6.5
caffe: 1.0
OpenCV：3.4.2
Anaconda3：5.2.0

您好！想用您转换的yolov5s网络在caffe上训练，要如何改动网络结构呢？

是训练单一目标检测，和您例子中的安全帽差不多，想问一下如何调整prototxt文件结构以开始训练过程呢？

boost::detail::set_tss_data 碰到找不到这个，请问作者你boost的版本是多少

.build_release/lib/libcaffe.so: undefined reference to boost::detail::set_tss_data(void const*, void (*)(void (*)(void*), void*), void (*)(void*), void*, bool)' collect2: error: ld returned 1 exit status Makefile:636: recipe for target '.build_release/tools/caffe.bin' failed make: *** [.build_release/tools/caffe.bin] Error 1 make: *** Waiting for unfinished jobs.... .build_release/lib/libcaffe.so: undefined reference to boost::detail::set_tss_data(void const*, void ()(void ()(void*), void*), void ()(void), void*, bool)'
collect2: error: ld returned 1 exit status
Makefile:636: recipe for target '.build_release/tools/upgrade_net_proto_binary.bin' failed
make: *** [.build_release/tools/upgrade_net_proto_binary.bin] Error 1
.build_release/lib/libcaffe.so: undefined reference to boost::detail::set_tss_data(void const*, void (*)(void (*)(void*), void*), void (*)(void*), void*, bool)' collect2: error: ld returned 1 exit status Makefile:636: recipe for target '.build_release/tools/upgrade_solver_proto_text.bin' failed make: *** [.build_release/tools/upgrade_solver_proto_text.bin] Error 1 .build_release/lib/libcaffe.so: undefined reference to boost::detail::set_tss_data(void const*, void ()(void ()(void*), void*), void ()(void), void*, bool)'
collect2: error: ld returned 1 exit status
Makefile:636: recipe for target '.build_release/tools/caffemodel2txt.bin' failed
make: *** [.build_release/tools/caffemodel2txt.bin] Error 1
.build_release/lib/libcaffe.so: undefined reference to boost::detail::set_tss_data(void const*, void (*)(void (*)(void*), void*), void (*)(void*), void*, bool)' collect2: error: ld returned 1 exit status Makefile:636: recipe for target '.build_release/tools/caffe_yolov5s.bin' failed make: *** [.build_release/tools/caffe_yolov5s.bin] Error 1 .build_release/lib/libcaffe.so: undefined reference to boost::detail::set_tss_data(void const*, void ()(void ()(void*), void*), void ()(void), void*, bool)'
collect2: error: ld returned 1 exit status
Makefile:636: recipe for target '.build_release/tools/extract_features.bin' failed
make: *** [.build_release/tools/extract_features.bin] Error 1
.build_release/lib/libcaffe.so: undefined reference to boost::detail::set_tss_data(void const*, void (*)(void (*)(void*), void*), void (*)(void*), void*, bool)' collect2: error: ld returned 1 exit status Makefile:636: recipe for target '.build_release/tools/compute_image_mean.bin' failed make: *** [.build_release/tools/compute_image_mean.bin] Error 1 .build_release/lib/libcaffe.so: undefined reference to boost::detail::set_tss_data(void const*, void ()(void ()(void*), void*), void ()(void), void*, bool)'
collect2: error: ld returned 1 exit status
Makefile:636: recipe for target '.build_release/tools/upgrade_net_proto_text.bin' failed
make: *** [.build_release/tools/upgrade_net_proto_text.bin] Error 1