Binary ground truth masks for the CASIA 2.0 image forensics dataset across various tampering categories. These reference images enable pixel-level localization and performance scoring for digital image forgery detection algorithms.
Use Cases
- Train image segmentation models to localize forgeries using the binary mask pixels as ground truth labels.
- Benchmark the F1-score and IoU of tampering detection algorithms on the CASIA 2.0 dataset.
- Analyze the localization error of forgery detection methods by comparing predicted heatmaps to these reference masks.
Strengths
- Binary masks representing tampered regions for pixel-level analysis.
- Direct mapping to the CASIA 2.0 image forgery dataset.
- Reference images for evaluating image manipulation detection.