tf-faster-rcnn GPU配置详细步骤

Environment

Ubuntu 16.04
NVIDIA GTX 1060
CUDA 9.0
CuDNN 7.1.4
python3

Installation

• Install opencv, cython, easydict

pip install opencv
pip install cython
pip install easydict

• Clone the repository

git clone https://github.com/endernewton/tf-faster-rcnn.git

• Build the cython modules

cd tf-faster-rcnn/lib
vim Makefile
#line 2：python -> python3

make clean
make
cd ..

• Install the Python COCO API

cd data
git clone https://github.com/pdollar/coco.git
cd coco/PythonAPI

vim Makefile
#line 2：python -> python3

make
cd ../..

• Setup data(e.g: PASCAL VOC 2007)

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

#VOC basic structure
$VOCdevkit/                           # development kit
$VOCdevkit/VOCcode/                   # VOC utility code
$VOCdevkit/VOC2007                    # image sets, annotations, etc.

cd $FRCN_ROOT/data                    
ln -s $VOCdevkit VOCdevkit2007        #$VOCdevkit: VOCdevkit的目录位置

• Download resnet101 for voc pretrained model on 07+12 dataset

baidu cloud : https://pan.baidu.com/s/1LXHHuyYRr0fetnhQtKBayw
tar xvf voc_0712_80k-110k.tgz

• Create a folder and a soft link to use the pre-trained model

NET=res101
TRAIN_IMDB=voc_2007_trainval+voc_2012_trainval
mkdir -p output/${NET}/${TRAIN_IMDB}
cd output/${NET}/${TRAIN_IMDB}                    #模型放在这个目录下面

ln -s ../../../data/voc_2007_trainval+voc_2012_trainval ./default  #建立一个软链接
cd ../../..

• Run the demo

GPU_ID=0
CUDA_VISIBLE_DEVICES=$GPU_ID 
python3 ./tools/demo.py                  


-------------------------------------------------------------------------------------------
#如果遇到错误No module named 'tkinter'
在相应的.py文件中import matplotlib.pyplot as plt之前加上
import matplotlib as mpl
mpl.use('Agg')

参考gpu服务器使用matplotlib错误：http://www.mamicode.com/info-detail-2190555.html

• Modify the program before train & test

vim ./lib/datasets/voc_eval.py
#line 121: 'w' -> 'wb'
#line 105: cachefile = os.path.join(cachedir, '%s_annots.pkl' % imagesetfile   
改为
cachefile = os.path.join(cachedir, '%s_annots.pkl' % imagesetfile.split("/")[-1].split(".")[0])

vim ./experiments/scripts/train_faster_rcnn.sh
#line 22: ITERS=70000 改为 ITERS=2000            #自己设置跑的轮数
#line 62: python -> python3
#line 73: python -> python3

vim ./experiments/scripts/test_faster_rcnn.sh
#line 22: ITERS=70000 改为 ITERS=2000
#line 58: python -> python3
#line 67: python -> python3

• Test with pretrained resnet101 model

rm data/VOCdevkit2007/VOC2007/ImageSets/Main/test.txt_annots.pkl
GPU_ID=0
./experiments/scripts/test_faster_rcnn.sh $GPU_ID pascal_voc_0712 res101

• Result

 

• Train your own model

1) Download pre-trained models and weights

#VGG16
mkdir -p data/imagenet_weights
cd data/imagenet_weights
wget -v http://download.tensorflow.org/models/vgg_16_2016_08_28.tar.gz
tar -xzvf vgg_16_2016_08_28.tar.gz
mv vgg_16.ckpt vgg16.ckpt
cd ../..

#ResNet101
mkdir -p data/imagenet_weights
cd data/imagenet_weights
wget -v http://download.tensorflow.org/models/resnet_v1_101_2016_08_28.tar.gz
tar -xzvf resnet_v1_101_2016_08_28.tar.gz
mv resnet_v1_101.ckpt res101.ckpt
cd ../..

2) Train & Test

#Train
./experiments/scripts/train_faster_rcnn.sh [GPU_ID] [DATASET] [NET]
# GPU_ID is the GPU you want to test on
# NET in {vgg16, res50, res101, res152} is the network arch to use
# DATASET {pascal_voc, pascal_voc_0712, coco} is defined in train_faster_rcnn.sh
# Examples:
./experiments/scripts/train_faster_rcnn.sh 0 pascal_voc vgg16
./experiments/scripts/train_faster_rcnn.sh 1 coco res101

#Test
./experiments/scripts/test_faster_rcnn.sh [GPU_ID] [DATASET] [NET]
# GPU_ID is the GPU you want to test on
# NET in {vgg16, res50, res101, res152} is the network arch to use
# DATASET {pascal_voc, pascal_voc_0712, coco} is defined in test_faster_rcnn.sh
# Examples:
./experiments/scripts/test_faster_rcnn.sh 0 pascal_voc vgg16
./experiments/scripts/test_faster_rcnn.sh 1 coco res101

参考：https://github.com/endernewton/tf-faster-rcnn