CoTracker: It is Better to Track Together

Meta AI Research, GenAI; University of Oxford, VGG

Nikita Karaev, Ignacio Rocco, Benjamin Graham, Natalia Neverova, Andrea Vedaldi, Christian Rupprecht

[Paper] [Project] [BibTeX]

CoTracker is a fast transformer-based model that can track any point in a video. It brings to tracking some of the benefits of Optical Flow.

CoTracker can track:

Every pixel in a video
Points sampled on a regular grid on any video frame
Manually selected points

Try these tracking modes for yourself with our Colab demo.

Installation Instructions

Ensure you have both PyTorch and TorchVision installed on your system. Follow the instructions here for the installation. We strongly recommend installing both PyTorch and TorchVision with CUDA support.

Steps to Install CoTracker and its dependencies:

git clone https://github.com/facebookresearch/co-tracker
cd co-tracker
pip install -e .
pip install opencv-python einops timm matplotlib moviepy flow_vis

Download Model Weights:

mkdir checkpoints
cd checkpoints
wget https://dl.fbaipublicfiles.com/cotracker/cotracker_stride_4_wind_8.pth
wget https://dl.fbaipublicfiles.com/cotracker/cotracker_stride_4_wind_12.pth
wget https://dl.fbaipublicfiles.com/cotracker/cotracker_stride_8_wind_16.pth
cd ..

Running the Demo:

Try our Colab demo or run a local demo with 10*10 points sampled on a grid on the first frame of a video:

python demo.py --grid_size 10

Evaluation

To reproduce the results presented in the paper, download the following datasets:

And install the necessary dependencies:

pip install hydra-core==1.1.0 mediapy tensorboard

Then, execute the following command to evaluate on BADJA:

python ./cotracker/evaluation/evaluate.py --config-name eval_badja exp_dir=./eval_outputs dataset_root=your/badja/path

Training

To train the CoTracker as described in our paper, you first need to generate annotations for Google Kubric MOVI-f dataset. Instructions for annotation generation can be found here.

Once you have the annotated dataset, you need to make sure you followed the steps for evaluation setup and install the training dependencies:

pip install pytorch_lightning==1.6.0

Now you can launch training on Kubric. Our model was trained for 50000 iterations on 32 GPUs (4 nodes with 8 GPUs).

python train.py --batch_size 1 --num_workers 28 \
--num_steps 50000 --ckpt_path ./ --model_name cotracker \
--save_freq 200 --sequence_len 24 --eval_datasets tapvid_davis_first badja \
--traj_per_sample 256 --sliding_window_len 8 --updateformer_space_depth 6 --updateformer_time_depth 6 \
--save_every_n_epoch 10 --evaluate_every_n_epoch 10 --model_stride 4

License

The majority of CoTracker is licensed under CC-BY-NC, however portions of the project are available under separate license terms: Particle Video Revisited is licensed under the MIT license, TAP-Vid is licensed under the Apache 2.0 license.

Acknowledgments

We would like to thank PIPs and TAP-Vid for publicly releasing their code and data. We also want to thank Luke Melas-Kyriazi for proofreading the paper, Jianyuan Wang, Roman Shapovalov and Adam W. Harley for the insightful discussions.

Citing CoTracker

If you find our repository useful, please consider giving it a star ⭐ and citing our paper in your work:

@article{karaev2023cotracker,
  title={CoTracker: It is Better to Track Together},
  author={Nikita Karaev and Ignacio Rocco and Benjamin Graham and Natalia Neverova and Andrea Vedaldi and Christian Rupprecht},
  journal={arXiv:2307.07635},
  year={2023}
}

Nangal / co-tracker