Towards Total Recall in Industrial Anomaly Detection

Roth, Karsten; Pemula, Latha; Zepeda, Joaquin; Schölkopf, Bernhard; Brox, Thomas; Gehler, Peter V.

doi:10.1109/cvpr52688.2022.01392

Cited by 369 publications

(287 citation statements)

References 25 publications

Supporting

Mentioning

285

Contrasting

Unclassified

Order By: Relevance

“…Note that we could reach the reported

AUC of PatchCore only in a single fold of cross-validation, but not on average over different folds of cross-validation and, therefore, report slightly lower average scores than in [ 22 ]. The same appears to be the case for CutPaste and we could only touch the reported

AUC.…”

Section: Local Anomaly Scorementioning

confidence: 67%

“…The required pre-training of the networks is always done by using the well-known ImageNet dataset, where all architectures reach a test set accuracy of around 85%. Note, we concatenated two adjacent layers

of EfficientNet-B4 as the layers are relatively low-dimensional, which improves the performance slightly (see, e.g., [ 22 ]). Due to the pooling layers, the spatial resolution of the feature map decreases with the depth of the network, i.e., deeper layers have lower spatial resolution.…”

Section: Methodsmentioning

confidence: 99%

“…Although the Deep SVD is a deep learning based version of the singular value decomposition (SVD), the DROCC method uses a nearest neighbor approach on pre-trained neural network features. For local defect detection, a recent method named PatchCore [ 22 ] achieves almost total recall on the MVTec challenge [ 7 ] using a greedy algorithm for dataset reduction based on coreset theory [ 23 ]. It is also based on an underlying nearest neighbor search in the feature space of a pre-trained EfficientNet-B4, but uses a modified distance as anomaly score.…”

Section: Related Workmentioning

confidence: 99%

See 2 more Smart Citations

Fast and Efficient Image Novelty Detection Based on Mean-Shifts

Hermann

Umlauf

Goldluecke

et al. 2022

Sensors

View full text Add to dashboard Cite

Image novelty detection is a repeating task in computer vision and describes the detection of anomalous images based on a training dataset consisting solely of normal reference data. It has been found that, in particular, neural networks are well-suited for the task. Our approach first transforms the training and test images into ensembles of patches, which enables the assessment of mean-shifts between normal data and outliers. As mean-shifts are only detectable when the outlier ensemble and inlier distribution are spatially separate from each other, a rich feature space, such as a pre-trained neural network, needs to be chosen to represent the extracted patches. For mean-shift estimation, the Hotelling T2 test is used. The size of the patches turned out to be a crucial hyperparameter that needs additional domain knowledge about the spatial size of the expected anomalies (local vs. global). This also affects model selection and the chosen feature space, as commonly used Convolutional Neural Networks or Vision Image Transformers have very different receptive field sizes. To showcase the state-of-the-art capabilities of our approach, we compare results with classical and deep learning methods on the popular dataset CIFAR-10, and demonstrate its real-world applicability in a large-scale industrial inspection scenario using the MVTec dataset. Because of the inexpensive design, our method can be implemented by a single additional 2D-convolution and pooling layer and allows particularly fast prediction times while being very data-efficient.

show abstract

“…Note that we could reach the reported

AUC.…”

Section: Local Anomaly Scorementioning

confidence: 67%

Section: Methodsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Fast and Efficient Image Novelty Detection Based on Mean-Shifts

Hermann

Umlauf

Goldluecke

et al. 2022

Sensors

View full text Add to dashboard Cite

show abstract

“…With the vigorous development of deep learning (DL) technology, 8 many scholars and institutions have studied anomaly detection methods based on DL, which can be roughly divided into reconstruction‐based 9–13 and pretrained networks 14–17 methods. In the reconstruction‐based method, only anomaly‐free samples are used for training in the training stage, and anomaly detection is performed in the inference stage according to the poor reconstruction effect for unknown anomaly regions.…”

Section: Introductionmentioning

confidence: 99%

“…Defard et al 15 used pretrained CNN to generate patch embedding vectors, exploited multivariate Gaussian distribution to obtain the probabilistic representation of the normal class, and used Mahalanobis distance to calculate anomaly scores of images to achieve anomaly localization. Roth et al 16 adopted a greedy coreset mechanism to reduce the memory bank extracted by the pretrained network, and then obtained anomaly scores according to the distance between the features in the memory bank and the patch features of the test image. Rudolph et al 17 proposed to input normal images of multiple scales into a pretrained feature extraction network, and used the normalizing flows to perform maximum likelihood training on the extracted features to obtain the distribution of normal images, and then calculated log‐likelihood to determine whether it is abnormal.…”

Section: Introductionmentioning

confidence: 99%

RDAD: A reconstructive and discriminative anomaly detection model based on transformer

Xie

Huang

Ning

et al. 2022

Int J of Intelligent Sys

View full text Add to dashboard Cite

Given the shortcomings of low detection accuracy and poor generalization performance of most current surface defect detection methods for industrial products, this paper proposes a reconstructive and discriminative anomaly detection model. The proposed method uses squeeze-and-excitation block to assign the attention of feature channels to enhance the sensitivity of related features and improve the ability of the model to learn normal and anomaly boundaries. In addition, channel transformer is introduced in the encoder-decoder, so that the decoder better fuses the features in the encoder and reduces the semantic gap, and enhances the segmentation ability of anomalous regions of the model. The model is only trained with normal samples, and completes the localization of anomalous regions while detecting anomalies. Experiments are conducted on the challenging MVTec anomaly detection and Magnetic Tile Defect data sets. Compared with the current state-of-the-art unsupervised anomaly detection methods, the model not only improves the accuracy of anomaly detection, but also has better generality.

show abstract

Synthetic Data Generation for Surface Defect Detection

et al. 2022

View full text Add to dashboard Cite

Ensuring continued quality is challenging, especially when the customer satisfaction is basically the provided service. It seems to become easier with new technologies like Artificial Intelligence. But to design an intelligent assistant, field data are necessary but not always available. Synthetic data are largely used to replace real data. Made with a Generative Adversarial Networks or a rendering engine, they aim to be as efficient as real ones to train a Neural Network. When synthetic data generation meet the challenge of object detection, its capacity to deal with defect detection challenge is unknown. Here we demonstrate how to generate these synthetic data to detect defects. Through iterations we apply different methods from literature to generate synthetic data for object detection. From how to extract a defect from the few data we have, to how to organize the scene before data synthesis. Our study suggests that defect detection may be performed by training an object detector neural network with synthetic data and gives a protocol to do so even if at this point, no field experiments have been conducted to verify our detector performances under real conditions. This experiment is the starting point to develop a mobile and automatic defect detector that might be adapted to ensure new product quality.

show abstract

Towards Total Recall in Industrial Anomaly Detection

Cited by 369 publications

References 25 publications

Fast and Efficient Image Novelty Detection Based on Mean-Shifts

Fast and Efficient Image Novelty Detection Based on Mean-Shifts

RDAD: A reconstructive and discriminative anomaly detection model based on transformer

Synthetic Data Generation for Surface Defect Detection

Contact Info

Product

Resources

About