Skip to content

[NeurIPS 2023] MixFormerV2: Efficient Fully Transformer Tracking

License

Notifications You must be signed in to change notification settings

MCG-NJU/MixFormerV2

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

36 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

MixFormerV2

The official implementation of the NeurIPS 2023 paper: MixFormerV2: Efficient Fully Transformer Tracking.

Model Framework

model

Distillation Training Pipeline

training

News

  • [Sep 22, 2023] MixFormerV2 is accpeted by NeurIPS 2023! 🎉

  • [May 31, 2023] We released two versions of the pretrained model, which can be accessed on Google Driver and NJU Box.

  • [May 26, 2023] Code is available now!

Highlights

✨ Efficient Fully Transformer Tracking Framework

MixFormerV2 is a well unified fully transformer tracking model, without any dense convolutional operation and complex score prediction module. We propose four key prediction tokens to capture the correlation between target template and search area.

✨ A New Distillation-based Model Reduction Paradigm

To further improve efficiency, we present a new distillation paradigm for tracking model, including dense-to-sparse stage and deep-to-shallow stage.

✨ Strong Performance and Fast Inference Speed

MixFormerV2 works well for different benchmarks and can achieve 70.6% AUC on LaSOT and 57.4% AUC on TNL2k, while keeping 165fps on GPU. To our best knowledge, MixFormerV2-S is the first transformer-based one-stream tracker which achieves real-time running on CPU.

Install the environment

Use the Anaconda

conda create -n mixformer2 python=3.6
conda activate mixformer2
bash install_requirements.sh

Data Preparation

Put the tracking datasets in ./data. It should look like:

   ${MixFormerV2_ROOT}
    -- data
        -- lasot
            |-- airplane
            |-- basketball
            |-- bear
            ...
        -- got10k
            |-- test
            |-- train
            |-- val
        -- coco
            |-- annotations
            |-- train2017
        -- trackingnet
            |-- TRAIN_0
            |-- TRAIN_1
            ...
            |-- TRAIN_11
            |-- TEST

Set project paths

Run the following command to set paths for this project

python tracking/create_default_local_file.py --workspace_dir . --data_dir ./data --save_dir .

After running this command, you can also modify paths by editing these two files

lib/train/admin/local.py  # paths about training
lib/test/evaluation/local.py  # paths about testing

Train MixFormerV2

Training with multiple GPUs using DDP. You can follow instructions (in Chinese now) in training.md. Example scripts can be found in tracking/train_mixformer.sh.

bash tracking/train_mixformer.sh

Test and evaluate MixFormerV2 on benchmarks

  • LaSOT/GOT10k-test/TrackingNet/OTB100/UAV123/TNL2k. More details of test settings can be found in tracking/test_mixformer.sh.
bash tracking/test_mixformer.sh

TODO

  • Progressive eliminating version of training.
  • Fast version of test forwarding.

Contant

Tianhui Song: [email protected]

Yutao Cui: [email protected]

Citiation

@misc{mixformerv2,
      title={MixFormerV2: Efficient Fully Transformer Tracking}, 
      author={Yutao Cui and Tianhui Song and Gangshan Wu and Limin Wang},
      year={2023},
      eprint={2305.15896},
      archivePrefix={arXiv}
}

About

[NeurIPS 2023] MixFormerV2: Efficient Fully Transformer Tracking

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published