DailyPapers (@HuggingPapers)
2026-01-03 | โค๏ธ 50 | ๐ 5
Diffusion Knows Transparency
Repurposes video diffusion for transparent object depth and normal estimation. DKT achieves zero-shot SOTA on ClearPose/DREDS, runs at 0.17s/frame (1.3B), and provides temporally consistent predictions for arbitrary-length videos. https://x.com/HuggingPapers/status/2007245681388294302/photo/1
๐ ์๋ณธ ๋งํฌ
๋ฏธ๋์ด
๐ Related
Auto-generated - needs manual review