Anti-UAV tracking competition requires algorithms to track a given UAV target and simultaneously estimate the tracking states of the target. When the target disappears, an invisible mark of the target needs to be given.


🔥🔥🔥 Codalab server is online now!🔥🔥🔥

Guideline for Challenge

The test-challenge will be released on 3rd July. The deadline for result submission is 10th July.

We build a baseline method on Github, Please refer to the evaluation code and resulting output file to test your algorithm and prepare final submission (.zip) to Codalab server.

If you encounter any questions or misunderstandings, please feel free to contact us.,,

We also set up a QQ group (907888704) for quick communication.

Participation requirements

The dataset of test-dev is NOT allowed to be used in learning.

Applying pretrained UAV detecter is NOT allowed.

Contestants are encouraged to submit previously published trackers or modified versions of third-party trackers.

The submission description should clearly state the algorithm framework.


Best Paper: Certificate + 500 USD + Gift

1st-Place: Certificate + 1500 USD + Gift (PaddlePaddle solutions get additional 2000 USD)

2nd-Place: Certificate + 1000 USD + Gift (PaddlePaddle solutions get additional 1500 USD)

3rd-Place: Certificate + 500 USD + Gift (PaddlePaddle solutions get additional 1000 USD)


The dataset consists of 280 high quality, Full HD Thermal Infrared video sequences, spanning multiple occurrences of multi-scale UAVs.


We define the tracking accuracy as:

For frame t, IoUt is Intersection over Union (IoU) between the predicted tracking box and its corresponding ground-truth box, pt equals 1 when the predicted box is empty and 0 otherwise, and vt is the ground-truth existence/visibility flag of the target. The Iverson bracket indicator function 1 [vt>0] equals 1 when [vt>0] and 0 otherwise. The accuracy is averaged over all T frames.

Results Format

For tracking with bounding boxes, please use the following format:







Note: box coordinates are floats measured from the top left image corner (and are 0-indexed). An empty list denotes there is no target in the current frame.

Baseline and Evaluation Code

Baseline models and Evaluation codes are available on Github.

We provide Thermal Infrared (IR) videos and their ground-truth labels. Contestants can only use the ground- truth bounding box of the target in the first frame. For a comprehensive comparison of different trackers, we evaluate them on video sequences with various challenging attributes. Our evaluation ranks are calculated according to the overall performance on all video sequences.