Yolact_minimal

This is the same implementation made here with just small adjustements to be used in Colab with a single GPU.

Environments

Works fine on Colab with T4 GPU.

Prepare

1. Create a colab netbook and make sure to adjuste the ressources to GPU. In a cell :

git clone https://github.com/mohcenaouadj/SSL-YOLACT/

2. Install dependencies

# Build cython-nms 
python setup.py build_ext --inplace

Also (if needed) :

! pip install tensorboardX TensorRT terminaltables onnxruntime-gpu -q

3. Modify self.data_root in 'res101_coco' in config.py according to your data folder.
4. Download weights.

Yolact trained weights.

Backbone	box mAP	mask mAP	number of parameters	Google Drive
PixPro-Resnet50	40.53	38.66	30.7 M	best_38.66_res50_pascal_12000.pth
VicRegL-Resnet50	28.55	29.89	30.7 M	best_29.89_res50_pascal_11000.pth

5. Detect

# To detect images, pass the path of the image folder, detected images will be saved in `results/images`.
python detect.py --best_38.66_res50_pascal_12000.pth --image=images

# To detect videos, pass the path of video, detected video will be saved in `results/videos`:
python detect.py --weight=weights/best_38.66_res50_pascal_12000.pth --video=videos/1.mp4
# Use --real_time to detect real-timely.
python detect.py --weight=weights/best_38.66_res50_pascal_12000.pth --video=videos/1.mp4 --real_time

6. Use tensorboard

tensorboard --logdir=tensorboard_log/res50_pascal

7. In case you want to retrain :

You can download the pascal segmentation bounderies dataset from here

!torchrun --nproc_per_node=1 --master_port=$((RANDOM)) train.py --cfg=res50_pascal

Note that the configuration here is for pascal voc dataset, for more details you should check the original repo.

Citation

@inproceedings{yolact-iccv2019,
  author    = {Daniel Bolya and Chong Zhou and Fanyi Xiao and Yong Jae Lee},
  title     = {YOLACT: {Real-time} Instance Segmentation},
  booktitle = {ICCV},
  year      = {2019},
}

@article{liu2021Swin,
  title={Swin Transformer: Hierarchical Vision Transformer using Shifted Windows},
  author={Liu, Ze and Lin, Yutong and Cao, Yue and Hu, Han and Wei, Yixuan and Zhang, Zheng and Lin, Stephen and Guo, Baining},
  journal={arXiv preprint arXiv:2103.14030},
  year={2021}
}