PyTorch implementation of "Reconstruction by inpainting for visual anomaly detection (RIAD)"

Concept ot the RIAD [1].

Disjoint masks with $k={2, 4}$.
First and second row shows the mask of $k=2$ and $k=4$ respectively.
Each column shows the $i$-$th$ mask for each cell size $k$.

• Normal (Good): 1
• Abnormal (Not-good): 0, 2, 3, 4, 5, 6, 7, 8, 9 (other than 1)

Loss convergence

Reconstruction

Anomaly detection performance (w/ validation set)

Result of training: result.json (w/ test set)

• PyTorch 1.11.0

[1] Vitjan Zavrtanik et al. "Reconstruction by inpainting for visual anomaly detection." Pattern Recognition, vol. 112, 2021.
[2] Taiki Inoue. MSGMS (python module).