AI BriefingORTUS AI Open-Sources RightWayUp: A 360-Degree Image Rotation Model with Apache-2.0 License
Key point
ORTUS AI has open-sourced RightWayUp, a 360-degree image rotation detection model under the Apache-2.0 license, which achieves 93.0% accuracy on held-out tests and demonstrates superior robustness against JPEG compression compared to existing models.
Details
ORTUS AI, a video analytics company, has released RightWayUp, a new open-source model for estimating image rotation angles across all 360 degrees. The model, available in six sizes from Pico to Max under the Apache-2.0 license, was developed to detect camera orientation issues in CCTV frames. In held-out tests, RightWayUp Max achieved 93.0% accuracy within 10 degrees, outperforming the Woehrer 2026 model's 88.4%. On the Woehrer 2026 COCO-based benchmark, RightWayUp Max scored 98.8% (five-seed mean) compared to Woehrer's 98.0%, and achieved perfect accuracy on RotBench. A key finding was that existing models like Woehrer 2026 suffer significant accuracy drops (from 98.0% to 30.2%) when images are compressed as JPEG q90, likely due to reliance on rotated JPEG grid artifacts. RightWayUp models showed minimal degradation in these conditions. The code and weights are available on GitHub, with engineering assistance noted from Claude and Codex.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Whiteboard Launches Open-Source VS Code Fork for Human-AI Collaborative Software Design
OpenComputer Hosts 'Ship Video' Agent Using Claude Opus 5.5 to Generate Explainer Videos from Code