SLSG: Industrial Image Anomaly Detection by Learning Better Feature Embeddings and One-Class Classification

Introduction

Industrial image anomaly detection under the setting of one-class classification has significant practical value. However, most existing models face challenges in extracting separable feature representations when performing feature embedding and in constructing compact descriptions of normal features when performing one-class classification. One direct consequence of this is that most models perform poorly in detecting logical anomalies which violate contextual relationships. Focusing on more effective and comprehensive anomaly detection, we propose a network based on self-supervised learning and self-attentive graph convolution (SLSG). SLSG uses a generative pre-training network to assist the encoder in learning the embedding of normal patterns and the reasoning of position relationships. Subsequently, SLSG introduces the pseudo-prior knowledge of anomaly through simulated abnormal samples. By comparing the simulated anomalies, SLSG can better summarize the normal patterns and narrow down the hypersphere used for one-class classification. In addition, with the construction of a more general graph structure, SLSG comprehensively models the dense and sparse relationships among elements in the image, which further strengthens the detection of logical anomalies. Extensive experiments on benchmark datasets show that SLSG achieves superior anomaly detection performance, demonstrating the effectiveness of our method.

Datasets

Note the change of dataset path in the program.

• MVTec LOCO AD
• Describable Textures Dataset (DTD)

To calculate the image-level ROAUC scores for the MVTec LOCO AD dataset, the LOCO AD dataset needs to be organized into the following format:

MVTec LOCO AD dataset
|—— breakfast_box
|   |—— ground_truth
|	|	|-- logical_anomalies
|   |   	|—— 000_mask.png
|   |   	|—— 001_mask.png
|	|	|-- structural_anomalies
|   |   	|—— 000_mask.png
|   |   	|—— 001_mask.png
|   |—— test
|	|	|-- good
|   |   	|—— 000.png
|   |   	|—— 001.png
|	|	|-- logical_anomalies
|   |   	|—— 000.png
|   |   	|—— 001.png
|	|	|-- logical_anomalies
|   |   	|—— 000.png
|   |   	|—— 001.png
|   |—— train
|	|	|-- good
|   |   	|—— 000.png
|   |   	|—— 001.png
|—— juice_bottle
|   |—— ...
|   |—— ...

Training

python run_training.py

Inference

python run_inference.py

We provide weights for model inference：https://drive.google.com/drive/folders/18VTqznST_fbebgtyu_w8nfbYvVFiNaYj?usp=drive_link

Image-level evaluation metrics (AUROC) of the provide model weights on the MVTec LOCO AD dataset:

	Breakfast box	Juice bottle	Pushpins	Screw bag	Splicing connector	Mean
Structural	0.829	0.977	0.941	0.932	0.935	0.923
Logical	0.944	0.991	0.967	0.734	0.862	0.899
Mean	0.884	0.985	0.955	0.808	0.894	0.905

Reference Codes

• https://github.com/LilitYolyan/CutPaste
• https://github.com/VitjanZ/DRAEM
• https://github.com/huawei-noah/Efficient-AI-Backbones/tree/master