Study: Optical Flow Datasets Saturated, RAFT Still Holds Up
kwangmoo_yi · x · 2026-07-15
The post shares a paper by Reijalt et al. on the real-world generalization of optical flow models. The author concludes that current optical flow datasets have reached saturation. In practical applications, if you are still using the classic RAFT model, you won't miss out on significant performance gains by not using the latest datasets.
Related event: Optical Flow Datasets Saturated: Classic RAFT Still Sufficient(2 posts)→
More from Research
- Anthropic masterclass spotlights how to build and observe AI agents — _jaydeepkarale · 2026-07-21
- NeurIPS 2026 workshop calls papers on on-device intelligence — YiMaTweets · 2026-07-21
- AI Security Institute says every tested model tried to cheat in cyber evaluations — connoraxiotes · 2026-07-21
- AI companies are buying old books to avoid training on AI-generated slop — CackleRooster · 2026-07-21
- Sakana says multiple diffusion models plus MCTS beat test-time scaling on coding and math — SakanaAILabs · 2026-07-21
- Soofi S 30B-A3B releases a full pretraining report and claims open-model leads in English and German — abursuc · 2026-07-21