Contamination Map Methodology

Vision / Model

APRIL-GAN (VAND 2023), trained on MVTec AD

Organisation
Zhejiang University et al. (Xuhai Chen, Yue Han, Jiangning Zhang)
Family
APRIL-GAN · version mvtec_pretrained.pth
Released
2023-05-27 (arXiv v1 submission date (https://arxiv.org/abs/2305.17382))
Checkpoints
exps/pretrained/mvtec_pretrained.pth
Weights
Open · MIT
Links
PaperCode

Training data disclosure: Full list

Level determined for 12% of benchmark cells.

“Conversely, when testing on VisA, we trained the linear layers using the test set of MVTec AD.”
arxiv.org · paper · Section 3.1 · retrieved 2026-10-15
“where `mvtec_pretrained.pth` represents the model trained on the MVTec AD dataset and `visa_pretrained.pth` represents the model trained on the VisA dataset.”
raw.githubusercontent.com · readme · Test / Pretrained Models · retrieved 2026-10-15

Built on

Everything these base models may have seen is inherited by APRIL-GAN (VAND 2023), trained on MVTec AD.

Other declared training data

Benchmark cells

Industrial anomaly detection

Test data seen MVTec AD · in fine-tuning data: its test split was used for training

Evidence (2)
“Moreover, as the training of the newly introduced linear layers relies on the ground-truth anomaly maps, we solely use the test set of the MVTec AD [2] dataset.”
arxiv.org · paper · Section 3.1 (VAND challenge zero-shot track) · retrieved 2026-10-15
“Conversely, when testing on VisA, we trained the linear layers using the test set of MVTec AD.”
arxiv.org · paper · Section 3.1 · retrieved 2026-10-15

Unknown MVTec AD 2 · unknown for its base model OpenAI CLIP: training data not disclosed in enough detail

Unknown MVTec LOCO AD · unknown for its base model OpenAI CLIP: training data not disclosed in enough detail

Unknown VisA · unknown for its base model OpenAI CLIP: training data not disclosed in enough detail

Unknown Real-IAD · unknown for its base model OpenAI CLIP: training data not disclosed in enough detail

Unknown MPDD · unknown for its base model OpenAI CLIP: training data not disclosed in enough detail

Unknown BTAD · unknown for its base model OpenAI CLIP: training data not disclosed in enough detail

Unknown KolektorSDD2 · unknown for its base model OpenAI CLIP: training data not disclosed in enough detail

Notes

Frozen CLIP ViT-L/14-336 (OpenAI) plus trained linear projection layers. This checkpoint (trained on the MVTec AD test set, 3 epochs, with 4-image mosaic augmentation) is used for zero-/few-shot results on VisA and for the VAND 2023 challenge submission (the challenge test set is a modified VisA). Few-shot mode reuses the same weights without further fine-tuning, plus a memory bank of k normal reference images.

Sources