Clear Memory-Augmented Auto-Encoder for Surface Defect Detection

Luo, W.; Niu, Tongzhi; Tang, Lixin; Yu, Wei; Li, Bin

doi:10.48550/arxiv.2208.03879

Cited by 1 publication

(4 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…They struggle to achieve noise-free background reconstructions and often reconstruct defects during testing. Memory-augmented autoencoders [31,42,43] were proposed to solve the partial reconstruction of defects. Memory-augmented autoencoder-based methods depend on restoring defects for inspection and often struggle to restore complex defects.…”

Section: Related Workmentioning

confidence: 99%

“…For the accurate simulation of real-world defect occurrences, artificial defect generation must carefully consider the target image's (I T ) specific details where defects will be introduced. Previous methods, such as Cutpaste [30] and CMA-AE [31], generate artificial anomalies by cropping a rectangular shape from a random source image and pasting it at a random position within a target image. Our approach involves cutting a portion from source images with varying shapes and sizes, and then pasting it randomly onto the target image as shown in Figure 2.…”

Section: Artificial Defect Generation Algorithm (Adga)mentioning

confidence: 99%

“…To validate the effectiveness of our proposed method, we compared its defect detection performance against those of several prominent anomaly and defect detection methods, including AE-SSIM [24], AnoGAN [23], f-AnoGAN [35], MS-FCAE [40], MemAE [42], TrustMAE [43], VAE [49], ACDN [50], AFEAN [29], and CMA-AE [31]. Table 1 summarizes the quantitative analysis results.…”

Section: Overall Performance Comparisonmentioning

confidence: 99%

“…CutPaste [30] uses a method that involves cropping a rectangular image patch and pasting it at a random location within a larger image. CMA-AE [31] uses a general artificial anomaly generation algorithm and involves cropping a rectangular area from a natural image in the ImageNet dataset [32] and pasting it at a random location within a normal training image. In our work, we propose an artificial defect generation algorithm (ADGA).…”

mentioning

confidence: 99%

See 3 more Smart Citations

Autoencoder-Based Unsupervised Surface Defect Detection Using Two-Stage Training

Getachew Shiferaw,

Yao

2024

J. Imaging

View full text Add to dashboard Cite

Accurately detecting defects while reconstructing a high-quality normal background in surface defect detection using unsupervised methods remains a significant challenge. This study proposes an unsupervised method that effectively addresses this challenge by achieving both accurate defect detection and a high-quality normal background reconstruction without noise. We propose an adaptive weighted structural similarity (AW-SSIM) loss for focused feature learning. AW-SSIM improves structural similarity (SSIM) loss by assigning different weights to its sub-functions of luminance, contrast, and structure based on their relative importance for a specific training sample. Moreover, it dynamically adjusts the Gaussian window’s standard deviation (σ) during loss calculation to balance noise reduction and detail preservation. An artificial defect generation algorithm (ADGA) is proposed to generate an artificial defect closely resembling real ones. We use a two-stage training strategy. In the first stage, the model trains only on normal samples using AW-SSIM loss, allowing it to learn robust representations of normal features. In the second stage of training, the weights obtained from the first stage are used to train the model on both normal and artificially defective training samples. Additionally, the second stage employs a combined learned Perceptual Image Patch Similarity (LPIPS) and AW-SSIM loss. The combined loss helps the model in achieving high-quality normal background reconstruction while maintaining accurate defect detection. Extensive experimental results demonstrate that our proposed method achieves a state-of-the-art defect detection accuracy. The proposed method achieved an average area under the receiver operating characteristic curve (AuROC) of 97.69% on six samples from the MVTec anomaly detection dataset.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Artificial Defect Generation Algorithm (Adga)mentioning

confidence: 99%