Constrained R-CNN: A general image manipulation detection model

Open paper
Benchmark Model Rank Results
image-manipulation-detection-on-casia-v1CR-CNN–Balanced Accuracy: .481AUC: .670
image-manipulation-detection-on-cocoglideCR-CNN–Balanced Accuracy: .447AUC: .589
image-manipulation-detection-on-columbiaCR-CNN–Balanced Accuracy: .631AUC: .755
image-manipulation-detection-on-coverageCR-CNN–Balanced Accuracy: .391AUC: .553
image-manipulation-detection-on-dso-1CR-CNN–Balanced Accuracy: .289AUC: .576
image-manipulation-localization-on-casia-v1CR-CNN–Average Pixel F1(Fixed threshold): .481
image-manipulation-localization-on-cocoglideCR-CNN–Average Pixel F1(Fixed threshold): .447
image-manipulation-localization-on-columbiaCR-CNN–Average Pixel F1(Fixed threshold): .631
image-manipulation-localization-on-coverageCR-CNN–Average Pixel F1(Fixed threshold): .391
image-manipulation-localization-on-dso-1CR-CNN–Average Pixel F1(Fixed threshold): .289

Rank counts only results with a code link.