SPot-the-Difference Self-Supervised Pre-training for Anomaly Detection and Segmentation

By Yang Zou, Jongheon Jeong, Latha Pemula, Dongqing Zhang, Onkar Dabeer. European Conference on Computer Vision, 2022

729 citations recorded by the source corpus.

Visual anomaly detection is commonly used in industrial quality inspection. In this paper, we present a new dataset as well as a new self-supervised learning method for ImageNet pre-training to improve anomaly detection and segmentation in 1-class and 2-class 5/10/high-shot training setups. We release the Visual Anomaly (VisA) Dataset consisting of 10,821 high-resolution color images (9,621 normal and 1,200 anomalous samples) covering 12 objects in 3 domains, making it the largest industrial anomaly detection dataset to date. Both image and pixel-level labels are provided. We also propose a new self-supervised framework - SPot-the-difference (SPD) - which can regularize contrastive self-supervised pre-training, such as SimSiam, MoCo and SimCLR, to be more suitable for anomaly detection tasks. Our experiments on VisA and MVTec-AD dataset show that SPD consistently improves these contrastive pre-training baselines and even the supervised pre-training. For example, SPD improves Area Under the Precision-Recall curve (AU-PR) for anomaly segmentation by 5.9% and 6.8% over SimSiam and supervised pre-training respectively in the 2-class high-shot regime. We open-source the project at http://github.com/amazon-research/spot-diff .

Subfields: Artificial Intelligence, Radiology, Nuclear Medicine and Imaging

View Paper: https://www.semanticscholar.org/paper/2c1837be2e4e6c4f47d51c42773e398897c372a6

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Guoyang Xie
2. Feng Zheng
3. Ming Tang
4. Shangwen Li
5. Yaochu Jin
6. Jingbao Wang
7. Chengjie Wang
8. Shibo He

Citations in the corpus

• AnomalyCLIP: Object-agnostic Prompt Learning for Zero-shot Anomaly Detection (2023, positive): “We use the top feature map as local visual details for anomaly segmentation.”
• Deep Industrial Image Anomaly Detection: A Survey (2023, positive): “Zou et al. [67] use contrastive learning to train the backbone network and propose a new data augmentation method called SPD to push the network to differentiate between two images with slight differences.”
• AnomalyGPT: Detecting Industrial Anomalies using Large Vision-Language Models (2023, methodology): “However, existing IAD datasets [1, 37] contain only a few thousand samples, rendering direct fine-tuning easy to overfitting and catastrophic forgetting.”
• WinCLIP: Zero-/Few-Shot Anomaly Classification and Segmentation (2023, neutral): “We see that Win-CLIP significantly improves over using these na¨ıve adaptations of CLIP on both MVTec-AD and VisA.”