A collection of PyQt5-based image labeling tools designed for image restoration tasks. Built to streamline quality assessment and hallucination detection in reconstructed images.
Three specialized labeling interfaces:
Binary classification tool for marking hallucinated regions in image reconstructions.
- Click to toggle severity of hallucinations
- Displays ground truth and hallucination proxy mask for reference
- Automatic progress saving
Assess overall reconstruction quality across multiple variants.
- Side-by-side comparison with ground truth
- Binary accept/reject labeling
- Resume from last labeled image
Rank multiple reconstructions by quality or hallucination severity.
- Drag-and-drop or click-based ranking (1-3 scale)
- Compare normal, intrinsic, and extrinsic hallucinations
- Color-coded feedback (Green=best, Orange=medium, Red=worst)
pip install PyQt5 numpy