Temporal Forgery Localization on LAV-DF cross-dataset evaluation from AV-Deepfake1M (test)
21.81mAP@IoU=0.1RT-DeepLoc
Evaluation results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| RT-DeepLocSupervision=weakly2026.01 | 21.81 | 20.36 | 18.39 | 16.63 | 14.88 | 13.47 | 11.06 | 16.66 | 82.41 | 82.41 | 82.39 | 77.55 | 81.19 | |
| UMMAFormerSupervision=fully2026.01 | 13.93 | 13.56 | 13.13 | 12.76 | 12.42 | 12.14 | 11.83 | 12.82 | 32.35 | 31.98 | 31.54 | 30.43 | 31.58 | |
| MFMSSupervision=fully2026.01 | 13.48 | 12.66 | 12.01 | 11.52 | 11.04 | 10.59 | 10.06 | 11.62 | 35.38 | 32.86 | 30.51 | 27.49 | 31.56 | |
| CoLASupervision=weakly2026.01 | 1.35 | 0.13 | 0.03 | 0.01 | 0.01 | 0.01 | 0.01 | 0.22 | 20.71 | 19.95 | 16.13 | 8.17 | 11.06 | |
| LOCOSupervision=weakly2026.01 | 0.27 | 0.04 | 0.01 | 0 | 0 | 0 | 0 | 0.05 | 5.95 | 5.95 | 5.72 | 3.54 | 5.29 |