At some point in the not-too-distant future, pairing depth-sensing cameras with artificial intelligence will enable computers to truly understand the physical world — a vision that researchers and companies have spent years working to transform into reality. Today, open source computer vision library OpenCV announced that it’s celebrating its 20th anniversary with a global competition designed to spur practical innovations in spatial AI, and will award both 1,200 depth sensing cameras and cash prizes to help entrants prototype their projects.
Backed by Intel and Microsoft Azure, the OpenCV AI Competition consists of two phases: Phase 1 requires a team to identify a real-world AI problem that can be solved over a three-month period with OpenCV’s latest neural inference depth camera, OAK-D. In Phase 2, each of the top 210 teams will receive either 4 or 10 OAK-D cameras, as well as 100 hours of free Azure NC6 processing time, access to Intel’s Dev Cloud, and OpenCV Slack support, all for the purpose of actually solving the identified AI problem. OpenCV will award $5,000-$20,000 prizes to the top three projects globally, plus three $2,000-$5,000 regional prizes across each of six competition regions, and $2,000 regional popular vote prizes.
The OpenCV competition is significant for technical decision makers because it offers enterprise and educational teams the opportunity to develop practical computer vision applications with hardware, software, and potentially monetary support from several of the CV industry’s most prominent organizations. For many teams, the competition will offer enough OAK-D hardware and Azure service to get an entire research group up and running on developing spatial AI solutions, with the prospect of winning a cash reward for particularly impressive results. Consequently, there may well be a contest-inspired burst of highly practical computer vision applications covering multiple geographies and industries.
OpenCV has identified six key project categories for the competition — visually impaired assistance, education, health and fitness, agriculture, COVID, and robotics — plus a catchall “miscellaneous” category for apps that don’t fit the other categories. The goal is to encourage ethically appropriate, non-discriminatory solutions across these categories, following TensorFlow’s and Microsoft’s responsible AI guidelines.
OAK-D combines a 12-megapixel RGB camera with a stereo depth engine containing hardware-accelerated AI and computer vision processors. The 4K camera runs at 30 frames per second without consuming processing time from the host computer’s CPU, capturing and processing spatial data that can augment Intel’s OpenVINO deep learning models. It’s capable of identifying objects and people in real time, subtracting backgrounds, and estimating motion, as well as supporting pose estimation and semantic segmentation; H265 video is also exported for easy visualization and analysis.
Timing is the competition’s only major hitch. Although OpenCV’s publicity campaign began today, the event’s Phase 1 submission deadline is coming up very quickly — January 27 — which means that unless the timeline for submissions is extended, interested teams will need to get their proposals together right away. OpenCV plans to select Phase 2 finalist teams on February 11, with a June 27 deadline for project completion, and winner selection by July 12. Teams interested in signing up can do so here.
VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact.
Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:
- up-to-date information on the subjects of interest to you
- our newsletters
- gated thought-leader content and discounted access to our prized events, such as Transform
- networking features, and more