Awesome

Feature-Distilled Transformer for UAV Tracking

Changhong Fu, Haobo Zuo, Guangze Zheng, Junjie Ye, and Bowen Li

Abstract

Occlusion, fast motion, and illumination variation are prone to cause object feature pollution, which lead to crucial problems for visual tracking, especially from unmanned aerial vehicle (UAV) perspectives in the intelligent transportation field. The key reason is that most trackers directly exploit search patch features for location estimation without considering if they are contaminated. To address this issue, this work proposes an efficient and effective feature-distilled Transformer (FDT) for aerial tracking, which can alleviate feature pollution by purifying search features with uncontaminated temporal information. FDT consists of two primary parts, a feature encoder to enhance internal attention of feature maps and a distillation decoder guided by the purifying strategy. Specifically, the distillation decoder fully exploits temporal information about feature pollution from the last frame, thereby efficiently guiding the distillation of feature maps. Consequently, pure features of the object are accurately reserved to improve the tracking performance in challenging UAV scenarios. Exhaustive experiments on four authoritative UAV tracking benchmarks have validated that FDT achieves the state-of-the-art performance, especially on sequences with severe feature pollution. In addition, FDT has strongly proved its practicability with 32.7 frames per second in the real-world tests on an aerial platform.

About Code

1. Environment setup

This code has been tested on Ubuntu 18.04, Python 3.8.3, Pytorch 0.7.0/1.6.0, CUDA 10.2. Please install related libraries before running this code:

  pip install -r requirements.txt

2. Test

Download pretrained model: FDTmodel(code:r0fy) and put it into tools/snapshot directory.

Download testing datasets and put them into test_dataset directory. If you want to test the tracker on a new dataset, please refer to pysot-toolkit to set test_dataset.

   python test.py 
        --dataset UAV123                #dataset_name
        --snapshot snapshot/FDTmodel.pth  # tracker_name

The testing result will be saved in the results/dataset_name/tracker_name directory.

3. Train

Prepare training datasets

Download the datasets:

VID

COCO

GOT-10K

LaSOT

Train a model

To train the FDT model, run train.py with the desired configs:

   cd tools
   python train.py

4. Evaluation

We provide the tracking results(code: 35rw) of UAV123@10fps, DTB70, UAVTrack112_L, and UAV123. If you want to evaluate the tracker, please put those results into results directory.

    python eval.py 	                          \
         --tracker_path ./results          \ # result path
         --dataset UAV123                  \ # dataset_name
         --tracker_prefix 'FDTmodel'   # tracker_name

5. Contact

If you have any questions, please contact me.

Haobo Zuo

Email: 1951684@tongji.edu.cn

Demo Video

Acknowledgement

The code is implemented based on pysot. We would like to express our sincere thanks to the contributors.