README.md

McADTR: Multi-class Anomaly Detection Transformer with Heterogeneous Knowledge Distillation

2022 AIR-DISCOVER Summer Research Project, Tsinghua University.

Pytorch implementation of McADTR.

docs

model: new_student.py
train student: new_train_student.py
train teacher: teacher_train.py
test: evaluate_1.py
pixel information test: evaluate_~.py

dataset format:
.../data/MVtec/
.../data/MAD1/bottle/.jpeg

Details

Acknowledgement

Modified based on the following code base：
1.[CVPR2020] Uninformed students student-teacher anomaly detection with discriminative latent embeddings
https://github.com/LuyaooChen/uninformed-students-pytorch
2.[ICCV2021] Learning mult-scene absolute pose regression with transformers
https://github.com/yolish/multi-scene-pose-transformer

Citation

@misc{weize2022airmcadtr,
title = {McADTR: Multi-class Anomaly Detection Transformer with Heterogeneous Knowledge Distillation},
author = {Weize Li, Qiang Zhou, Hao Zhao},
journal = {AIR-DISCOVER Project},
url = {https://github.com/EricLee0224/McADTR},
year = {2022}
}