Change Detection on OSCD 30 (val)
74.7PrecisionSeCo
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SeCoSupervision=S (Self-supervision only)2021.12 | 74.7 | 15.2 | 25.26 | |
| U-NetSupervision=F (Full Supervision), Weight Initialization=random2021.12 | 70.53 | 19.17 | 29.44 | |
| U-NetSupervision=F (Full Supervision), Weight Initialization=ImageNet2021.12 | 70.42 | 25.12 | 36.2 | |
| SeCoSupervision=S+F (Self-supervised pre-training followed by fully supervised fine-tuning)2021.12 | 65.47 | 38.06 | 46.94 | |
| MoCo-v2Supervision=S+F (Self-supervised pre-training followed by fully supervised fine-tuning)2021.12 | 64.49 | 30.94 | 40.71 | |
| OursSupervision=S+F (Self-supervised pre-training followed by fully supervised fine-tuning), Mode=fine-tuned2021.12 | 61.8 | 57.13 | 59.37 | |
| DeepLab-v3Supervision=F (Full Supervision), Weight Initialization=ImageNet2021.12 | 51.63 | 51.06 | 53.04 | |
| OursSupervision=S (Self-supervision only)2021.12 | 37.52 | 72.65 | 49.48 | |
| MoCo-v2Supervision=S (Self-supervision only)2021.12 | 29.21 | 11.92 | 16.93 | |
| VCASupervision=S (Self-supervision only)2021.12 | 9.92 | 20.77 | 13.43 |