Video Class Agnostic Segmentation

[Method Paper] [Benchmark Paper] [Project] [Demo]

Official Datasets and Implementation from our Paper "Video Class Agnostic Segmentation Benchmark in Autonomous Driving" in Workshop on Autonomous Driving, CVPR 2021.

Installation

This repo is tested under Python 3.6, PyTorch 1.4

• Download Required Packages

pip install -r requirements.txt
pip install "git+https://github.com/cocodataset/panopticapi.git"

• Setup mmdet

python setup.py develop

Motion Segmentation Track

Dataset Preparation

• Follow Dataset Preparation Instructions.

Inference

• Download Trained Weights on Ego Flow Suppressed, trained on Cityscapes and KITTI-MOTS

• Modify Configs according to dataset path + Image/Annotation/Flow prefix

configs/data/kittimots_motion_supp.py
configs/data/cscapesvps_motion_supp.py

• Evaluate CAQ,

python tools/test_eval_caq.py CONFIG_FILE WEIGHTS_FILE

CONFIG_FILE: configs/infer_kittimots.py or configs/infer_cscapesvps.py

• Qualitative Results

python tools/test_vis.py CONFIG_FILE WEIGHTS_FILE --vis_unknown --save_dir OUTS_DIR

• Evaluate Image Panoptic Quality, Note: evaluated on 1024x2048 Images

python tools/test_eval_ipq.py configs/infer_cscapesvps_pq.py WEIGHTS_FILE --out PKL_FILE

Training

Coming Soon ...

Open-set Segmentation Track

Coming soon ...

Acknowledgements

Dataset and Repository relied on these sources:

• Voigtlaender, Paul, et al. "Mots: Multi-object tracking and segmentation." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2019.
• Kim, Dahun, et al. "Video panoptic segmentation." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020.
• Wang, Xinlong, et al. "Solo: Segmenting objects by locations." European Conference on Computer Vision. Springer, Cham, 2020.
• This Repository built upon SOLO Code

Citation

@article{siam2021video,
      title={Video Class Agnostic Segmentation Benchmark for Autonomous Driving}, 
      author={Mennatullah Siam and Alex Kendall and Martin Jagersand},
      year={2021},
      eprint={2103.11015},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

Contact

If you have any questions regarding the dataset or repository, please contact [email protected].