Multi-feature Reconstruction Network using Crossed-mask Restoration for Unsupervised Industrial Anomaly Detection

TL;DR

This work converts the image reconstruction into a combination of parallel feature restorations and proposes a multi-feature reconstruction network, MFRNet, using crossed-mask restoration, which is highly competitive with or significantly outperforms other state-of-the-arts on four public available datasets and one self-made dataset.

Abstract

Unsupervised anomaly detection using only normal samples is of great significance for quality inspection in industrial manufacturing. Although existing reconstruction-based methods have achieved promising results, they still face two problems: poor distinguishable information in image reconstruction and well abnormal regeneration caused by model under-regularization. To overcome the above issues, we convert the image reconstruction into a combination of parallel feature restorations and propose a multi-feature reconstruction network, MFRNet, using crossed-mask restoration in this paper. Specifically, a multi-scale feature aggregator is first developed to generate more discriminative hierarchical representations of the input images from a pre-trained model. Subsequently, a crossed-mask generator is adopted to randomly cover the extracted feature map, followed by a restoration network based on the transformer structure for high-quality repair of the missing regions. Finally, a hybrid loss is equipped to guide model training and anomaly estimation, which gives consideration to both the pixel and structural similarity. Extensive experiments show that our method is highly competitive with or significantly outperforms other state-of-the-arts on four public available datasets and one self-made dataset.

Figures (11)

• Figure 1: (a) Reconstruction model on image space tends to yield good reconstruction for anomalies using limited semantic information. (b) Our method reconstructs multi-scale features through a masking and restoration model for accurate anomaly detection.
• Figure 2: An overview of our unsupervised anomaly detection pipeline.
• Figure 3: Example of masking operation with parameters $k=2$ and $n=3$ on a single-channel feature map. The masked regions within each subset are complementary, thus covering all possible anomalies.
• Figure 4: (a) Architecture of the proposed restoration network. (b) Stucture of the hybrid transformer block.
• Figure 5: The proposed acquisition equipment for fabric image.