faster-rcnn.pytorch项目环境配置（从0到1）

faster-rcnn.pytorch项目环境配置（从0到1）

其实pytorch版本和CUDA版本高，都没有关系！！！都可以适配，显卡30系、20系都没关系，都可以用！

下面我将在AutoDL平台上，从0到1安装配置一遍！！！

1. conda create -n DA python=3.8
2. conda activate DA
3. 克隆代码

git clone https://github.com/jwyang/faster-rcnn.pytorch.git
cd faster-rcnn.pytorch
git checkout pytorch-1.0

4. pip install -r requirements.txt

若安装opencv时卡死，请单独安装opencv-python，并附带版本:

pip install opencv-python==3.4.9.31

5. 安装pytorch

conda install pytorch==1.11.0 torchvision==0.12.0 torchaudio==0.11.0 cudatoolkit=11.3 -c pytorch

6. 编译coco api

cd lib/pycocotools #进入coco文件夹
新建 setup.py 文件，并将以下内容复制

# --------------------------------------------------------
# Fast R-CNN
# Copyright (c) 2015 Microsoft
# Licensed under The MIT License [see LICENSE for details]
# Written by Ross Girshick
# --------------------------------------------------------

import os
from setuptools import setup
from distutils.extension import Extension
from Cython.Distutils import build_ext
import numpy as np


# Obtain the numpy include directory.  This logic works across numpy versions.
try:
   numpy_include = np.get_include()
except AttributeError:
   numpy_include = np.get_numpy_include()


ext_modules = [
   Extension(
       '_mask',
       sources=['maskApi.c', '_mask.pyx'],
       include_dirs = [numpy_include, '.'],
       extra_compile_args=['-Wno-cpp', '-Wno-unused-function', '-std=c99'],
   )
]

setup(
   name='faster_rcnn',
   ext_modules=ext_modules,
   # inject our custom trigger
   cmdclass={'build_ext': build_ext},
)

随后python setup.py build_ext --inplace

7. 编译cuda依赖

cd …
python setup.py build develop # 编译cuda 依赖项

若发生以下错误：

pytorch版本问题：pytorch在最新的版（1.11）本中将THC/THC.h文件删除了。降低pytorch版本即可
conda install pytorch==1.10.0 torchvision==0.11.0 torchaudio==0.10.0 cudatoolkit=11.3 -c pytorch -c conda-forge

8. 在根目录下创建data，下载数据集/预训练权重

mkdir data
cd data

wget http://host.robots.ox.ac.uk/pascal/VOC/voc2007/VOCtrainval_06-Nov-2007.tar
wget http://host.robots.ox.ac.uk/pascal/VOC/voc2007/VOCtest_06-Nov-2007.tar
wget http://host.robots.ox.ac.uk/pascal/VOC/voc2007/VOCdevkit_08-Jun-2007.tar

tar xvf VOCtrainval_06-Nov-2007.tar
tar xvf VOCtest_06-Nov-2007.tar
tar xvf VOCdevkit_08-Jun-2007.tar

# 在xxx/data路径下
ln -s /root/autodl-tmp/faster-rcnn.pytorch/data/VOCdevkit VOCdevkit2007

mkdir pretrained_model
cd pretrained_model
wget https://filebox.ece.vt.edu/~jw2yang/faster-rcnn/pretrained-base-models/resnet101_caffe.pth
wget https://filebox.ece.vt.edu/~jw2yang/faster-rcnn/pretrained-base-models/vgg16_caffe.pth

9. 更改代码：

autodl-tmp/faster-rcnn.pytorch/lib/roi_data_layer/minibatch.py
第15行： from scipy.misc import imread
改为： from imageio import imread  
pip install imageio

autodl-tmp/faster-rcnn.pytorch/lib/model/utils/config.py
第374行：yaml_cfg = edict(yaml.load(f))
改为：yaml_cfg = edict(yaml.safe_load(f))

随后

CUDA_VISIBLE_DEVICES=0 python trainval_net.py --dataset pascal_voc --net vgg16 --epochs 20 --bs 1 --nw 2 --lr 1e-2 --lr_decay_step 8 --use_tfb --cuda

参考：

1

2

[3](
https://blog.csdn.net/cube_suger/article/details/112648721