Workshop and Competition on RISEx 2025
Workshop and Competition on RISEx 2025
We are excited to invite students and postdoctoral fellows to participate in the Alberta RISEx 2025 competition! This is an excellent opportunity to showcase your skills in AI and computer vision for robotics, collaborate with peers, and win awards.
Overview
Unmanned Aerial Vehicles (UAVs), as intelligent aerial robots that integrate perception and decision-making, play a vital role in a wide range of applications, including agricultural management, disaster relief, and urban surveillance. Leveraging UAVs for object localization, tracking, and segmentation enables effective monitoring of plant and animal growth, supports search and rescue operations, and facilitates the understanding and recognition of target behaviors. The competition will focus on how to build an efficient and robust model for UAV-based object tracking and segmentation using visible and thermal data.
Competition Details
The competition consists of two tracks. For each track, a development set will be released for training and testing purposes. All competitors will be required to evaluate their models on a private test set, where only the initial bounding boxes will be provided. The two tracks are independent. Participants are welcome to participate in either or both.
Visual Object Tracking Track: Taking the initial bounding box as input, the model is required to generate the bounding boxes of the target in all the frames. Success Rate and Precision Rate are used as metrics.
Video Object Segmentation Track: Taking the initial mask as input, the model should output the pixel-level binary mask to indicate the target location and shape. J&F score is used for evaluation.
News
[Aug. 1] The website for MTSUAV@RISEx2025 is open.
[Aug. 24] A submission template can be found in the website.
[Aug. 30] Data is available at [VOT Track] [VOS Track].
[Aug. 30] An Introduction on data usage is available here.
[Sep. 17] If you are interested in participating, please send an email to pz9@ualberta.ca with MTSUAV@RISEx2025 in the subject line so we can count the number of participants and prepare the competition schedule. Thanks for your cooperation.
[Oct. 28] Our workshop will be held at 3:15 – 4:45 PM, November 14, 2025. Welcome to Join!
Workshop Schedule
[3:15–3:45 PM] Report on the MTSUAV Competition -- Dr. Pengyu Zhang (UoA)
[3:45–4:05 PM] Invited Tutorial on Multi-modal Visual Tracking -- Dr. Pengyu Zhang (UoA)
[4:05–4:20 PM] Networking and Break
[4:20–4:40 PM] Invited Talk on RGB-Thermal Video Semantic Segmentation: Recent Advances and Trends -- Dr. Jingjing Li (UoA)
Important Date
• Aug. 25th 2025 – Competition Open. • Nov. 1st 2025 – Winners Announcement.
• Oct. 10th 2025 – Submission Deadline. • Nov. 13th 2025 – Winners Talk.
Eligibility
Teams can consist of up to 3 members. Both undergraduates, graduate students, and postdoctoral fellows are welcome.
At least one team member must be registered for the conference to represent the team.
Key Requirements
Submission rules: Each team is required to submit a single ZIP archive via email
(mtsuav.risex2025@gmail.com).
Submission format: The archive should contain the model files, a model description and result files on the test set. A submission template can be found here
Model files: Both runnable code and weights should be submitted.
Model description: The description should contain the model framework, key design and strengths, etc. (maximum 300 word with figures).
Result files: For VOT track, each file (in TXT format) represents the results of one video and each line in the TXT file represents the target’s position in a frame as (x_min, y_min, w, h). For VOS track, a folder consisting all the binary masks of the whole video, with the same file name of image files.
Real-time speed: All models must run at real-time speed on a consumer-grade GPU (i.e., > 20 FPS on a computer with an AMD Ryzen 7965WX CPU and NVIDIA RTX 4090 GPU), demonstrating their potential for deployment on cutting-edge devices.
Forbidden dataset for training: Models are allowed to train on any public datasets EXCEPT the provided test set.
Contact
Dr. Pengyu Zhang (pz9@ualberta.ca) Dr. Bita Fallahi (fallahi@ualberta.ca)