If you are interested in ML model serving, feel free to have a look at my another project on model serving framework!
A Deep Attentive Contour Model for Efficient Instance Segmentation (PDF)
DANCE's Pipeline | Illustration |
The codes in master
branch are mainly for experiments on COCO; if you want to have a look on the codes for experiments on SBD / Cityscapes, welcome to checkout the snake
branch, which is developed based on the codebase of previous art.
conda create --name dance python==3.6.8
conda deactivate && conda activate dance
conda install pytorch==1.4.0 torchvision==0.5.0 cudatoolkit=10.1 -c pytorch
git clone https://github.com/lkevinzc/dance && cd dance && pip install -r requirements.txt && cd ..
git clone https://github.com/facebookresearch/detectron2.git && cd detectron2 && git checkout 1a7daee064eeca2d7fddce4ba74b74183ba1d4a0 && python -m pip install -e . && cd ..
cd dance/core/layers/extreme_utils && export CUDA_HOME="/usr/local/cuda-10.1" && python setup.py build_ext --inplace
pip install cython; pip install 'git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI'
fvcore
version: pip install fvcore==0.1.1.dev200512
dance/datasets/coco
model name | AP | AP50 | AP75 | weights |
---|---|---|---|---|
dance_r50_3x | 36.8 | 58.5 | 39.0 | link |
dance_r101_3x | 38.1 | 60.2 | 40.5 | link |
note: put them under output/
python train_net.py --config-file configs/Dance_R_50_3x.yaml --eval-only MODEL.WEIGHTS ./output/r50_3x_model_final.pth
python train_net.py --config-file configs/Dance_R_101_3x.yaml --eval-only MODEL.WEIGHTS ./output/r101_3x_model_final.pth
Any discussion or suggestion is welcome! Feel free to contact the author via liuzichen@u.nus.edu
:)
If you find this project helpful for your research, please consider citing using BibTeX below:
@InProceedings{liu2021dance,
author = {Liu, Zichen and Liew, Jun Hao and Chen, Xiangyu and Feng, Jiashi},
title = {DANCE: A Deep Attentive Contour Model for Efficient Instance Segmentation},
booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
month = {January},
year = {2021},
pages = {345-354}
}