AllClip Inc. Wins Third Place Worldwide in “Spiideo SoccerNet SynLoc 2026” Competition at CVPR

AllClip Inc. (Headquarters: Chiyoda-ku, Tokyo; Representative Director: Ikuma Uchida; hereinafter “AllClip”) won third place worldwide in “Spiideo SoccerNet SynLoc 2026,” a competition held as part of CVPR 2026, one of the largest international conferences in computer vision.

CVPR is one of the world’s leading international conferences in computer vision and one of the largest academic gatherings for top researchers in image recognition technologies. Held alongside CVPR every year, the “SoccerNet Challenge” is a competition focused on AI processing accuracy for soccer video. In “Spiideo SoccerNet SynLoc 2026,” one of this year’s tracks, 85 teams from around the world participated, and our team achieved third place worldwide.

Competition Overview

“Spiideo SoccerNet SynLoc 2026” is a task that detects players from a single frame of soccer footage and identifies their positions in world coordinates. This is an important elemental technology for building video recognition and understanding AI that automates sports analysis. The evaluation design emphasizes not only player detection accuracy, but also positional accuracy on the field.

In this effort, we achieved results that significantly exceeded the baseline method by building a module that effectively detects small objects such as players in 4K images, using player pose information, and tuning player positions while considering camera parameters.

Future Outlook

By applying the methods used in this competition, we aim to further improve the accuracy of our AI camera system “AllClip,” add new features, and adapt the technology in ways that can be directly applied to challenges in sports settings.

Beyond this competition, we expect that contributing to the development of video AI technologies for recognizing and understanding sports scenes will help realize a future where value previously available only in top-level environments, such as extracting individual player scenes and measuring physical data, can be delivered to every category.

We will continue improving sports video analysis technologies and camera systems to transform every sports scene into a video and data experience, and we will promote practical AI sports systems that are easy to use in competitive settings.

If you are interested in AllClip, please feel free to contact us.