Skip to content

Latest commit

 

History

History
107 lines (72 loc) · 2.89 KB

README.md

File metadata and controls

107 lines (72 loc) · 2.89 KB
Raw

OPNet: Deep Occlusion Perception Network with Boundary Awareness for Amodal Instance Segmentation

This is the PyTorch implementation of OPNet built on the open-source detectron2.

"OPNet: Deep Occlusion Perception Network with Boundary Awareness for Amodal Instance Segmentation"

Installation

conda create -n opnet python=3.7 -y
source activate opnet
 
pip install torch==1.11.0+cu113 torchvision==0.12.0+cu113 -f https://download.pytorch.org/whl/torch_stable.html
 

pip install ninja yacs cython matplotlib tqdm
pip install opencv-python==4.4.0.40
pip install scikit-image
export INSTALL_DIR=$PWD
 
cd $INSTALL_DIR
git clone https://github.com/cocodataset/cocoapi.git
cd cocoapi/PythonAPI
python setup.py build_ext install
 
cd $INSTALL_DIR
git clone https://github.com/ZitengXue/OPNet.git
cd OPNet/
pip install -r requirements.txt
python3 setup.py build develop

Dataset Preparation

D2SA dataset

The D2S Amodal dataset could be found at mvtec-d2sa.

KINS dataset

Download the Images from KITTI dataset.

The Amodal Annotations could be found at KINS dataset

CWALT dataset

Follow WALT,we will release soon.

Expected folder structure for each dataset

$PROJECT_DIR/
|OPNet/
|-detectron2/
|-- data/
|---- datasets/
|------- KINS/
|---------- train_imgs/
|---------- test_imgs/
|---------- annotations/
|------------- train.json
|------------- test.json
|------- D2SA/
|-------- annotations/
|-------- images/
|------- CWALT/
|...

Then, See here for more details on data registration

Multi-GPU Training and evaluation on Validation set

cd OPNet

For example, on the D2SA dataset, we show how to train our method and other baselines.

1.Train our model (ResNet50 backbone) on D2SA dataset :

CUDA_VISIBLE_DEVICES=0,1 python tools/train_net.py --config-file configs/D2SA-AmodalSegmentation/bounday.yaml--num-gpus 2

Test

If you want to eval your saved checkpoints:

python tools/train_net.py --config-file configs/{your_yaml_file} 
--eval-only MODEL.WEIGHTS {your_OUTPUT_DIR}/model_final.pth' --num-gpus x

Acknowledgement

Our work benefits a lot from BCNet and VRSP-Net. Thanks for their wonderful works.