Edge Detection on NYUD Crispness-emphasized evaluation - CEval v2 (val)
77.5ODS ScoreRankED + MATCHED
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| RankED + MATCHEDInput modality=RGB, Training regime=Supervised, Post-processing=MATCHED2026.02 | 77.5 | 78.4 | 0.806 | 0.886 | |
| Diff.Edge + MATCHEDInput modality=RGB, Training regime=Supervised, Post-processing=MATCHED2026.02 | 75.9 | 76.2 | 0.703 | 0.937 | |
| SAUGE + MATCHEDInput modality=RGB, Training regime=Zero-shot, Train Dataset=BSDS, Post-processing=MATCHED2026.02 | 75.8 | 77.9 | 0.813 | 0.852 | |
| PiDiNet + MATCHEDInput modality=RGB, Training regime=Supervised, Post-processing=MATCHED2026.02 | 73.6 | 74.9 | 0.777 | 0.93 | |
| Diff.EdgeInput modality=RGB, Training regime=Supervised2026.02 | 73.2 | 73.8 | — | 0.846 | |
| Diff.Edge*Input modality=RGB, Training regime=Supervised, Training Source=Our training2026.02 | 72.7 | 73.3 | 0.693 | 0.863 | |
| SAUGE†Input modality=RGB, Training regime=Zero-shot, Train Dataset=BSDS, Checkpoint=Official2026.02 | 49.1 | 51.7 | 0.554 | 0.187 | |
| RankEDInput modality=RGB, Training regime=Supervised2026.02 | 47.7 | 49 | 0.483 | 0.146 | |
| EDTERInput modality=RGB, Training regime=Supervised2026.02 | 43 | 45.7 | — | 0.195 | |
| BDCNInput modality=RGB, Training regime=Supervised2026.02 | 42.6 | 45 | 0.362 | 0.162 | |
| PiDiNetInput modality=RGB, Training regime=Supervised2026.02 | 39.9 | 42.4 | — | 0.173 | |
| RCFInput modality=RGB, Training regime=Supervised2026.02 | 39.8 | 41.3 | — | — | |
| HEDInput modality=RGB, Training regime=Supervised2026.02 | 38.7 | 40.4 | — | — |