Track-Anything is a flexible and interactive tool for video object tracking and segmentation. It is developed upon Segment Anything can specify anything to track and segment via user clicks only. During tracking users can flexibly change the objects they wanna track or correct the region of interest if there are any ambiguities. These characteristics enable Track-Anything to be suitable for:
- Video object tracking and segmentation with shot changes.
- Visualized development and data annnotation for video object tracking and segmentation.
- Object-centric downstream video tasks such as video inpainting and editing.
🚀 Updates
-
2023/04/25: We are delighted to introduce Caption-Anything ✍️ an inventive project from our lab that combines the capabilities of Segment Anything Visual Captioning and ChatGPT.
-
2023/04/20: We deployed [DEMO] on Hugging Face 🤗!
Demo
Multiple Object Tracking and Segmentation (with XMem)
Video Object Tracking and Segmentation with Shot Changes (with XMem)
Video Inpainting (with E2FGVI)
Get Started
Linux & Windows
# Clone the repository:
git clone https://github.com/gaomingqi/Track-Anything.git
cd Track-Anything
# Install dependencies:
pip install -r requirements.txt
# Run the Track-Anything gradio demo.
python app.py --device cuda:0
# python app.py --device cuda:0 --sam_model_type vit_b # for lower memory usage
Annotated VOTS2023 dataset
Running the demo
# Clone the repository:
git clone -b beta https://github.com/gaomingqi/Track-Anything.git
# server 110
python app_vots.py --device cuda:3 --port 12221 --mask_save True --votdir /nvme-ssd/lizhe/dataset/vots2023 --sequence ants1
# server 108
python app_vots.py --device cuda:3 --port 12221 --mask_save True --votdir /ssd2/tracking/vots2023 --sequence ants1
Operation
check the sequence list in /vots/group_seq to find your sequence list
Press the Get-Video-Info button to initialize the sequence
Press the Tracking button to generate mask for each frames
Refine the mask by select the Image Selection button and refine the mask for tracking
Mask save path:
# server 110
/nvme-ssd/lizhe/dataset/vots2023/gt_mask/
# server 108
/ssd2/tracking/vots2023/gt_mask/
Citation
If you find this work useful for your research or applications please cite using this BibTeX:
@misc{yang2023track
title={Track Anything: Segment Anything Meets Videos}
author={Jinyu Yang and Mingqi Gao and Zhe Li and Shang Gao and Fangjing Wang and Feng Zheng}
year={2023}
eprint={2304.11968}
archivePrefix={arXiv}
primaryClass={cs.CV}
}
Acknowledgements
The project is based on Segment Anything XMem and E2FGVI. Thanks for the authors for their efforts.