Pixel-level Manipulation Detection on DEFACTO 12k
61.8F1 ScoreManTra-Net
Evaluation results
| Method | Links | |
|---|---|---|
| ManTra-NetThresholding strategy=Optimal threshold per model & testset, Training dataset=Other (Not CASIAv2)2021.04 | 61.8 | |
| ManTra-NetThreshold Strategy=Optimal threshold per model & testset, Training Data=Private2021.12 | 61.8 | |
| MVSS-NetThresholding strategy=Optimal threshold per model & testset, Training dataset=CASIAv22021.04 | 57.2 | |
| MVSS-NetThreshold Strategy=Optimal threshold per model & testset, Training Data=CASIAv22021.12 | 57.2 | |
| SPANThreshold Strategy=Optimal threshold per model & testset, Training Data=Private2021.12 | 57.1 | |
| MVSS-Net++Threshold Strategy=Optimal threshold per model & testset, Training Data=CASIAv22021.12 | 55.6 | |
| CAT-NetThreshold Strategy=Optimal threshold per model & testset, Training Data=Private2021.12 | 44.1 | |
| FCNThresholding strategy=Optimal threshold per model & testset, Training dataset=CASIAv22021.04 | 40.1 | |
| FCNThreshold Strategy=Optimal threshold per model & testset, Training Data=CASIAv22021.12 | 40.1 | |
| GSR-NetThresholding strategy=Optimal threshold per model & testset, Training dataset=CASIAv22021.04 | 37.9 | |
| GSR-NetThreshold Strategy=Optimal threshold per model & testset, Training Data=CASIAv22021.12 | 37.9 | |
| CR-CNNThresholding strategy=Optimal threshold per model & testset, Training dataset=CASIAv22021.04 | 34 | |
| CR-CNNThreshold Strategy=Optimal threshold per model & testset, Training Data=CASIAv22021.12 | 34 | |
| ManTra-NetThresholding strategy=Fixed threshold (0.5), Training dataset=Other (Not CASIAv2)2021.04 | 15.5 | |
| ManTra-NetThreshold Strategy=Fixed threshold (0.5), Training Data=Private2021.12 | 15.5 | |
| MVSS-NetThresholding strategy=Fixed threshold (0.5), Training dataset=CASIAv22021.04 | 13.7 | |
| MVSS-NetThreshold Strategy=Fixed threshold (0.5), Training Data=CASIAv22021.12 | 13.7 | |
| HP-FCNThresholding strategy=Optimal threshold per model & testset, Training dataset=Other (Not CASIAv2)2021.04 | 13.6 | |
| HP-FCNThreshold Strategy=Optimal threshold per model & testset, Training Data=Private2021.12 | 13.6 | |
| CR-CNNThresholding strategy=Fixed threshold (0.5), Training dataset=CASIAv22021.04 | 13.2 | |
| CR-CNNThreshold Strategy=Fixed threshold (0.5), Training Data=CASIAv22021.12 | 13.2 | |
| FCNThresholding strategy=Fixed threshold (0.5), Training dataset=CASIAv22021.04 | 13 | |
| FCNThreshold Strategy=Fixed threshold (0.5), Training Data=CASIAv22021.12 | 13 | |
| H-LSTMThreshold Strategy=Optimal threshold per model & testset, Training Data=Published but no longer publicly accessible data2021.12 | 12.5 | |
| MVSS-Net++Threshold Strategy=Fixed threshold (0.5), Training Data=CASIAv22021.12 | 9.5 | |
| H-LSTMThreshold Strategy=Fixed threshold (0.5), Training Data=Published but no longer publicly accessible data2021.12 | 5.9 | |
| HP-FCNThresholding strategy=Fixed threshold (0.5), Training dataset=Other (Not CASIAv2)2021.04 | 5.5 | |
| HP-FCNThreshold Strategy=Fixed threshold (0.5), Training Data=Private2021.12 | 5.5 | |
| GSR-NetThresholding strategy=Fixed threshold (0.5), Training dataset=CASIAv22021.04 | 5.1 | |
| GSR-NetThreshold Strategy=Fixed threshold (0.5), Training Data=CASIAv22021.12 | 5.1 | |
| SPANThreshold Strategy=Fixed threshold (0.5), Training Data=Private2021.12 | 4.8 | |
| CAT-NetThreshold Strategy=Fixed threshold (0.5), Training Data=Private2021.12 | 4.6 |