The official PyTorch implementation of our CVPR 2023 Highlight paper:
Autoregressive Visual Trecking
GitHub maintainer: Yifan Bai
We present ARTrack, an autoregressive framework for visual object tracking. ARTrack tackles tracking as a coordinate sequence interpretation task that estimates object trajectories progressively, where the current estimate is induced by previous states and in turn affects subsequences. This time-autoregressive approach models the sequential evolution of trajectories to keep tracing the object across frames, making it superior to existing template matching based trackers that only consider the per-frame localization accuracy. ARTrack is simple and direct, eliminating customized localization heads and post-processings. Despite its simplicity, ARTrack achieves state-of-the-art performance on prevailing benchmark datasets.
| Variant | ARTrack-256 | ARTrack-384 | ARTrack-L-384 |
|---|---|---|---|
| Model Config | ViT-B, 256^2 resolution | ViT-B, 384^2 resolution | ViT-L, 384^2 resolution |
| GOT-10k (AO / SR 0.5 / SR 0.75) | 73.5 / 82.2 / 70.9 | 75.5 / 84.3 / 74.3 | 78.5 / 87.4 / 77.8 |
| LaSOT (AUC / Norm P / P) | 70.4 / 79.5 / 76.6 | 72.6 / 81.7 / 79.1 | 73.1 / 82.2 / 80.3 |
| TrackingNet (AUC / Norm P / P) | 84.2 / 88.7 / 83.5 | 85.1 / 89.1 / 84.8 | 85.6 / 89.6 / 84.8 |
| LaSOT_ext (AUC / Norm P / P) | 46.4 / 56.5 / 52.3 | 51.9 / 62.0 / 58.5 | 52.8 / 62.9 / 59.7 |
| TNL-2K (AUC) | 57.5 | 59.8 | 60.3 |
| NfS30 (AUC) | 64.3 | 66.8 | 67.9 |
| UAV123 (AUC) | 67.7 | 70.5 | 71.2 |
Our baseline model (backbone: ViT-B, resolution: 256x256) can run at 26 fps (frames per second) on a single NVIDIA GeForce RTX 3090, our alter decoder version can run at 45 fps on a single NVIDIA GeForce RTX 3090.
You can download the model weights and raw_result from Google Drive
| Variant | ARTrack-256 |
|---|---|
| Model Config | ViT-B, 256^2 resolution |
| GOT-10k (AO / SR 0.5 / SR 0.75) | 76.7 / 85.7 / 74.8 |
| LaSOT (AUC / Norm P / P) | 70.8 / 79.6 / 76.3 |
| TrackingNet (AUC / Norm P / P) | 84.3 / 88.7 / 83.4 |
| LaSOT_ext (AUC / Norm P / P) | 48.4 / 57.7 / 53.7 |
Use the Anaconda (CUDA 11.3)
conda env create -f ARTrack_env_cuda113.yaml
Run the following command to set paths for this project
python tracking/create_default_local_file.py --workspace_dir . --data_dir ./data --save_dir ./output
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
Put the tracking datasets in ./data. It should look like this:
${PROJECT_ROOT}
-- data
-- lasot
|-- airplane
|-- basketball
|-- bear
...
-- got10k
|-- test
|-- train
|-- val
-- coco
|-- annotations
|-- images
-- trackingnet
|-- TRAIN_0
|-- TRAIN_1
...
|-- TRAIN_11
|-- TEST
Download pre-trained MAE ViT-Base weights and put it under $PROJECT_ROOT$/pretrained_models (different pretrained models can also be used, see MAE for more details).
Since sequence-level training requires video input, and the COCO dataset contains only images, traditional training methods were first used to train the model so that it could be fairly compared to other trackers.
python tracking/train.py --script artrack --config artrack_256_full --save_dir ./output --mode multiple --nproc_per_node 4 --use_wandb 0
Replace --config with the desired model config under experiments/artrack. We use wandb to record detailed training logs, in case you don't want to use wandb, set --use_wandb 0.
To enable sequence-level training, replace 'experience/artrack_seq/*.yaml' PRETRAIN_PTH in the yaml configuration file with the path to your pretrained checkpoint, such as './output/artrack_256_full/checkpoints/train/artrack/artrack_256_full/ARTrack_ep0240.pth.tar'.
python tracking/train.py --script artrack_seq --config artrack_seq_256_full --save_dir ./output --mode multiple --nproc_per_node 4 --use_wandb 0
Change the corresponding values of lib/test/evaluation/local.py to the actual benchmark saving paths
Some testing examples:
- LaSOT or other off-line evaluated benchmarks (modify
--datasetcorrespondingly)
python tracking/test.py artrack_seq artrack_seq_256_full --dataset lasot --threads 16 --num_gpus 4
python tracking/analysis_results.py # need to modify tracker configs and names
- GOT10K-test
python tracking/test.py artrack_seq artrack_seq_256_full --dataset got10k_test --threads 16 --num_gpus 4
python lib/test/utils/transform_got10k.py --tracker_name ostrack --cfg_name vitb_384_mae_ce_32x4_got10k_ep100
- TrackingNet
python tracking/test.py artrack_seq artrack_seq_256_full --dataset trackingnet --threads 16 --num_gpus 4
python lib/test/utils/transform_trackingnet.py --tracker_name ostrack --cfg_name vitb_384_mae_ce_32x4_ep300
❤️❤️❤️Our idea is implemented base on the following projects. We really appreciate their excellent open-source works!
❤️❤️❤️This project is not for commercial use. For commercial use, please contact the author.
❤️❤️❤️This project is not for commercial use. For commercial use, please contact the author.
❤️❤️❤️This project is not for commercial use. For commercial use, please contact the author.
If any parts of our paper and code help your research, please consider citing us and giving a star to our repository.
@InProceedings{Wei_2023_CVPR,
author = {Wei, Xing and Bai, Yifan and Zheng, Yongchao and Shi, Dahu and Gong, Yihong},
title = {Autoregressive Visual Tracking},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2023},
pages = {9697-9706}
}
If you have any questions or concerns, feel free to open issues or directly contact me through the ways on my GitHub homepage provide below paper's title.