ATR Challenge

The Automatic Target Recognition Algorithm Challenge (ATR Challenge) focuses on those challenging problems in the field of target detection and recognition technology, releases application scenarios and datasets, and solicits target detection, tracking and recognition algorithms and application solutions.
The ATR Challenge 2025 will consist of two tracks, requiring participating teams to complete Track 1 and Track 2 synchronously. These two Tracks would be conducted based on the Automatic Target Recognition Scientific Data Open Platform (www.atrdata.cn). The scores from the two tracks will be normalized to 50 each, and their sum will be the participant's final score.

Track 1: Cross-Modal Visual Prompt-Based Object Matching and Recognition
Using visual prompt images containing specific target features, achieve efficient detection and precise localization of the target in images to be recognized across different modalities. Note: Visual prompt images and the images to be recognized may exhibit multi-dimensional differences, such as varying spectral bands, platforms, sources, resolutions, or perspectives.

Track 2: Optical Detection and Recognition of Vehicles in UAV Imagery
Addressing the challenges of low-altitude perception by unmanned aerial vehicles (UAVs) in complex urban environments, this track focuses on the precise detection and highly robust recognition of vehicle targets in visible-light images, despite variations in categories, viewpoints, scales, lighting, and weather conditions. The goal is to achieve accurate identification across multiple scenarios and conditions.

Team member and registration requirements
A team can have a maximum of 5 members (including 1 team leader) and 1 coach. The team members can be composed of graduate students, young teachers or researchers from global research institutions. The team leader can log into the website (www.atrdata.cn) to complete the registration.

Stage1: Online competition
The participating teams complete the algorithm and software design based on the public data set (including data set description and scoring rules) published on the competition website, submit to the competition platform and complete automatic testing and scoring based on the test data set (the test data set is not public during the competition, and the score list will be updated in real time), and the organizing committee selects 10 teams that enter the on-site defense stage based on the online competition ranking.

Stage2: On-site defense
Domestic teams in China are required to attend the on-site oral defense, while international teams unable to attend in person may submit a pre-recorded video presentation of their defense content. The Organizing Committee will determine final rankings by comprehensively evaluating online competition results and the performance during oral defense. Materials to be submitted at this stage include program reports (including schematic block diagrams of the algorithms used, theoretical models, algorithm steps, and result displays), pictures, algorithm source codes and executable programs, report ppts or videos, and other electronic materials.

Stage 3:Session presentation
PRIP (https://prip.by/2025/) will host an ATR Challenge Session, the winning participating teams are required to present and discuss their innovative solutions.

The prize details are subject to final confirmation, but the preliminary structure is as follows:
① First Prize (1 Winner)
② Second Prize (2 Winners)
③ Third Prize (3 Winners)
④ Outstanding Prizes (4 Winners)

Each winner will receive a bonus (the specific amount is still to be determined and will vary) along with a certificate and a trophy. Additionally, all participants who advance to the on-site defense stage and complete the competition will be eligible to receive a winning prize.

Paper Publication
After being reorganized and reviewed according to the requirements, the works entering the final stage can be recommended by the organizing committee to be published in the form of academic papers at the international conferences and academic journals, such as "IET Image Processing", "Drones", "Chinese Journal of Information Fusion", "Journal of System Simulation".