Overview
The OpenCV AI Competition 2026, powered by Amazon Web Services (AWS), invites teams to turn ambitious ideas into working computer-vision applications and share them with the global OpenCV community. The competition focuses on Physical AI and Generative AI systems in which visual understanding informs decisions, actions, predictions, or interactions. Every entry must use OpenCV 5 for substantive image or video analysis and run a meaningful component on AWS. Projects may use any programming language and supporting hardware.
The competition is sponsored by AWS and administered by the OpenCV Foundation. It is online and open to people above the legal age of majority in their country of residence, with listed country and territory exclusions. The organizers especially welcome applications in areas including autonomous inspection, process control, visual SLAM, digital twins, healthcare, accessibility, agriculture, environmental monitoring, smart cities, education, retail, and sports analytics.
Schedule and format
Registration and proposal submissions open during the competition launch period beginning August 12, 2026, with proposals accepted starting August 13. The two-month build phase runs from August 26 through October 26. Teams develop, evaluate, document, and submit their projects during this period. Final submissions are due October 26, 2026, at 11:59 p.m. Pacific Time. Final judging is scheduled for October 27 through November 9, with winners expected to be announced November 10 during an OpenCV Live! webinar or through another announced OpenCV channel.
Teams are encouraged to collaborate through the official Slack channel and share milestones using #OpenCVComp26. Teams awarded an AWS compute grant must attend a 30-minute Zoom check-in with an organizing committee member between October 7 and 14, maintain active development, and provide a concise progress update to receive the remaining half of the grant.
Project paths and submissions
Participants may enter either or both optional featured paths, or pursue another eligible approach. The Best Use of COOL path focuses on the Cloud-Optimized OpenCV Library running the core workload on AWS Graviton or the Arm portion of a documented hybrid system. Submissions should provide reproducible measurements and compare performance with an appropriate baseline. The Agentic Vision path requires a multi-step perception, decision, and action workflow in which OpenCV 5 results change a later plan, tool call, action, or request for human approval. A chatbot that merely explains a fixed vision result does not meet this path’s requirements.
Final entries must include a technical report, judge-accessible code, pinned dependencies and build instructions, an architecture diagram, a working endpoint or arranged live demonstration, and a judge-accessible video of no more than five minutes. Teams should also provide evaluation evidence, including relevant failure cases or limitations. Grant proposals ask for a team name, problem and intended impact, planned OpenCV 5 and AWS design, target users, evaluation and demonstration plans, chosen focus path, and team biographies.
Awards and judging
The listed total prize pool is $20,250. Awards include $5,000 for first place, $3,000 for second, and $2,000 for third. The COOL and Agentic Vision awards each offer $1,000. AWS compute grants are also listed at $150 each. Judges include Phil Nelson, OpenCV Director of Content & Creative, and Gary Bradski, OpenCV founder. Judging weighs technical execution (30%), innovation (20%), real-world impact (20%), user experience (10%), documentation and presentation (10%), and cloud delivery, reproducibility, and responsible operation (10%).
