Mask R-CNN 配置說明

概述：

本文主要介紹Mask R-CNN的配置和測試需要配置CUDA8.0、cuDNN6.0.21、Python2、Caffe2、Detectron(CPU支援不完善)、COCO API。

Caffe2安裝（cuda和cudnn略）

依賴關係：

for Ubuntu 14.04

sudo apt-get install -y --no-install-recommends libgflags2

for Ubuntu 16.04

sudo apt-get install -y --no-install-recommends libgflags-dev
 

for both Ubuntu 14.04 and 16.04

 sudo apt-get install -y --no-install-recommends \
       libgtest-dev \
       libiomp-dev \
       libleveldb-dev \
       liblmdb-dev \
       libopencv-dev \
       libopenmpi-dev \
       libsnappy-dev \
       openmpi-bin \
       openmpi-doc \
       python-pydot
 sudo pip install \
       flask \
       future \
       graphviz \
       hypothesis \
       jupyter \
       matplotlib \
       pydot python-nvd3 \
       pyyaml \
       requests \
       scikit-image \
       scipy \
       setuptools \
       six \
       tornado
 

下載並編譯caffe2

git clone --recursive https://github.com/caffe2/caffe2.git
cd caffe2

編譯

make
cd build && make install DESTDIR=/path/to/install

注：其中/path/to/install為安裝路徑，有sudo許可權可使用sudo make install 可安裝到/usr/local/下

驗證是否安裝成功

python -c 'from caffe2.python import core' 2>/dev/null && echo "Success" || echo "Failure"
 

輸出Success即成功

驗證GPU構建

python2 -c 'from caffe2.python import workspace; print(workspace.NumCudaDevices())'

輸出GPU數量即成功

加入環境變數

echo $PYTHONPATH
# export PYTHONPATH=/path/to/install/caffe2:$PYTHONPATH  #make install的路徑
# export PYTHONPATH=$PYTHONPATH: /path/to/caffe2/build  #Caffe2原始碼路徑
echo $LD_LIBRARY_PATH
# export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH

另外的依賴

Python依賴

pip install numpy pyyaml matplotlib opencv-python>=3.0 setuptools Cython mock

COCO API

# COCOAPI=/path/to/clone/cocoapi
git clone https://github.com/cocodataset/cocoapi.git $COCOAPI
cd $COCOAPI/PythonAPI
# 安裝到全域性第三方庫中
make install
# Alternatively, if you do not have permissions or prefer
# 安裝到使用者Python庫中
python2 setup.py install --user

Detectron安裝

下載Detectron原始碼

# DETECTRON=/path/to/clone/detectron
git clone https://github.com/facebookresearch/detectron $DETECTRON
Set up Python modules:

編譯：

cd $DETECTRON/lib && make

測試是否安裝成功

python2 $DETECTRON/tests/test_spatial_narrow_as_op.py

輸出如下即成功

Ran 3 tests in 7.049s
OK

使用Mask R-CNN演示

使用本地影象執行效果演示，使用tools/infer_simple.py，此例使用ResNet-101-FPN的Mask R-CNN作為模型。

python2 tools/infer_simple.py \
 --cfg configs/12_2017_baselines/e2e_mask_rcnn_R-101-FPN_2x.yaml \
 --output-dir /tmp/detectron-visualizations \
 --image-ext jpg \
 --wts https://s3-us-west2.amazonaws.com/detectron/35861858/12_2017_baselines/e2e_mask_rcnn_R-101-FPN_2x.yaml.02_32_51.SgT4y1cO/output/train/coco_2014_train:coco_2014_valminusminival/generalized_rcnn/model_final.pkl \
 demo

注：–cfg為模型路徑，–output-dir為輸出路徑 預設為’/tmp/infer_simple’，結果將會以pdf的形式輸出在指定目錄下；–image-ext為圖片格式，預設為jpg格式；–wts為網路權重檔案；demo為最後一個引數，為圖片所在資料夾路徑。

Mask R-CNN測試

使用來源於model zoo的end2end Mask R-CNN網路進行COCO_2014_minival中圖片的測試

python2 tools/test_net.py \
    --cfg configs/12_2017_baselines/e2e_mask_rcnn_R-101-FPN_2x.yaml \
    --multi-gpu-testing \
    TEST.WEIGHTS https://s3-us-west-2.amazonaws.com/detectron/35861858/12_2017_baselines/e2e_mask_rcnn_R-101-FPN_2x.yaml.02_32_51.SgT4y1cO/output/train/coco_2014_train:coco_2014_valminusminival/generalized_rcnn/model_final.pkl \
    NUM_GPUS 2

NUM_GPUS為使用GPU數量，測試結果如下

INFO json_dataset_evaluator.py: 232: ~~~~ Summary metrics ~~~~
  Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.364
  Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.585
  Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.387
  Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.166
  Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.392
  Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.540
  Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.303
  Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.459
  Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.478
  Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.266
  Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.520
  Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.642
 INFO json_dataset_evaluator.py: 122: Wrote json eval results to:./test/coco_2014_minival/generalized_rcnn/segmentation_results.pkl
 INFO task_evaluation.py:  65: Evaluating segmentations is done!
 INFO task_evaluation.py: 180: copypaste: Dataset: coco_2014_minival
 INFO task_evaluation.py: 182: copypaste: Task: box
 INFO task_evaluation.py: 185: copypaste: AP,AP50,AP75,APs,APm,APl
 INFO task_evaluation.py: 186: copypaste: 0.4089,0.6193,0.4478,0.2350,0.4421,0.5389
 INFO task_evaluation.py: 182: copypaste: Task: mask
 INFO task_evaluation.py: 185: copypaste: AP,AP50,AP75,APs,APm,APl
 INFO task_evaluation.py: 186: copypaste: 0.3639,0.5846,0.3869,0.1664,0.3915,0.5400

參考連結：