segmentation

Segmentation with SCL

This page provides basic tutorials about the usage of MMDetection with SCL.

Current code is develop on mmdetection@618dca08.

Setup

1. Install mmdet according to this doc, and run python setup.py develop in our repository.
2. Prepare datasets according to this.
3. Create a directory checkpoints/ in this folder and download pretrained models R-50-C4 caffe 1x and X-101-32x4d-FPN pytorch 1x to checkpoints/.

Train a model

Train with multiple GPUs

./tools/dist_train.sh ${CONFIG_FILE} ${GPU_NUM} [optional arguments]

More specifically, R-50-C4 can be trained with this command:

./tools/dist_train.sh configs/scl/mask_rcnn_r50_caffe_c4_1x.py 4 --validate

And X-101-32x4d-FPN can be trained with this command:

./tools/dist_train.sh configs/scl/mask_rcnn_x101_32x4d_fpn_1x.py 4 --validate

Notice

• Current code isn't compatible with basic models, so training base models (e.g. R-50-C4 without SCL) with this code is not feasible.
• We trained with 4 GPUs and other configuration hasn't been tried.
• Evaluating is similar to the above operations.

Help

Feel free to open an issue if you encounter troubles.

Citation

If you use this codebase in your research, please cite our paper:

@InProceedings{Zhou_2020_CVPR,
author = {Zhou, Mohan and Bai, Yalong and Zhang, Wei and Zhao, Tiejun and Mei, Tao},
title = {Look-Into-Object: Self-Supervised Structure Modeling for Object Recognition},
booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2020}
}