Estimating the Probability Density Function of New Fabrics for Fabric Anomaly Detection

Rippel, Oliver; Müller, Maximilian J.; Münkel, Andreas; Gries, Thomas; Merhof, Dorit

doi:10.5220/0010163604630470

Cited by 4 publications

(12 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Apart from the unimodal setting, AD may also occur in a multi-modal context [74], [75]. Here, less constrained methods are required to estimate the PDF of multi-modal normal data, such as the approach proposed by [45].…”

Section: Discussionmentioning

confidence: 99%

Gaussian Anomaly Detection by Modeling the Distribution of Normal Data in Pretrained Deep Features

Rippel

Mertens

Konig

et al. 2021

IEEE Trans. Instrum. Meas.

Self Cite

View full text Add to dashboard Cite

Anomaly Detection (AD) in images is a fundamental computer vision problem and refers to identifying images that deviate significantly from normality. State-of-the-art AD algorithms commonly learn a model of normality from scratch using task specific datasets in either semi-supervised or selfsupervised manner. We follow an alternative approach, and model the distribution of normal data in deep feature representations learned from ImageNet via a multivariate Gaussian (MVG). This lightweight approach achieves a new state of the art in AD on the public MVTec AD dataset. In addition to the empirical benefits, we give a clear motivation for the seemingly simplistic approach via the ties between deep generative and discriminative modeling revealed recently. We further elucidate why ImageNet representations are discriminative in the transfer learning AD setting using Principal Component Analysis. Here, we find that the principal components containing little variance in normal data are the ones crucial for discriminating between normal and anomalous instances, giving an explanation for the unreasonable effectiveness of our approach. We also investigate setting the working point of our approach by selecting acceptable False Positive Rate thresholds based on the MVG assumption as well as the resistance of our approach to unlabeled anomalies in the dataset. Finally, we investigate whether our approach is prone to exploiting spurious correlations using explainable AI techniques. Code is publicly available at https://github.com/ORippler/gaussian-ad-mvtec.

show abstract

Section: Discussionmentioning

confidence: 99%

Gaussian Anomaly Detection by Modeling the Distribution of Normal Data in Pretrained Deep Features

Rippel

Mertens

Konig

et al. 2021

IEEE Trans. Instrum. Meas.

Self Cite

View full text Add to dashboard Cite

show abstract

“…As an alternative, one could also try to leverage features yielded by either object detection or segmentation networks, which have been shown to generate features that maintain stronger spatial acuity [33], to improve AS performance. Last, anomalies may also occur in the multi-modal setting [43], and our current fine-tuning procedure can not be applied here. Here, less constrained priors such as Gaussian Mixture Models or Normalizing Flows may be used.…”

Section: Discussionmentioning

confidence: 99%

“…Furthermore, k-NN [12] has also been used to realize AD/AS on pretrained features. Last, generative algorithms such as Gaussian AD [11,49,14,43] are also commonly employed.…”

Section: Learning Anomaly Detection From Scratchmentioning

confidence: 99%

Transfer Learning Gaussian Anomaly Detection by Fine-tuning Representations

Rippel,

Chavan,

Lei

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Current state-of-the-art Anomaly Detection (AD) methods exploit the powerful representations yielded by large-scale ImageNet training. However, catastrophic forgetting prevents the successful fine-tuning of pre-trained representations on new datasets in the semi/unsupervised setting, and representations are therefore commonly fixed.In our work, we propose a new method to fine-tune learned representations for AD in a transfer learning setting. Based on the linkage between generative and discriminative modeling, we induce a multivariate Gaussian distribution for the normal class, and use the Mahalanobis distance of normal images to the distribution as training objective. We additionally propose to use augmentations commonly employed for vicinal risk minimization in a validation scheme to detect onset of catastrophic forgetting.Extensive evaluations on the public MVTec AD dataset reveal that a new state of the art is achieved by our method in the AD task while simultaneously achieving AS performance comparable to prior state of the art. Further, ablation studies demonstrate the importance of the induced Gaussian distri-

show abstract

“…Moreover, supervised algorithms generally outperform their semi-supervised counterparts [18,19]. However, supervised methods suffer from a major drawback: They generalize poorly to fabrics unseen during model training [20,21] and therefore do not meet the industrial requirement for low changeover costs. Instead, defective and defect-free data must be collected and annotated for every new fabric, which is a tedious, time-consuming, and expensive process.…”

Section: Introductionmentioning

confidence: 99%

“…While algorithms have been proposed to tackle this limitation, current research focuses on adapting converged models to new fabrics in a post hoc manner [20,21]. It thereby disregards the potential of training models that generalize better to unseen fabrics in the first place (see Figure 1).…”

Section: Introductionmentioning

confidence: 99%

Increasing the Generalization of Supervised Fabric Anomaly Detection Methods to Unseen Fabrics

Rippel

Zwinge

Merhof

2022

Sensors

Self Cite

View full text Add to dashboard Cite

Fabric ad tries to detect anomalies (i.e., defects) in fabrics, and fabric ad approaches are continuously improved with respect to their ad performance. However, developed solutions are known to generalize poorly to previously unseen fabrics, posing a crucial limitation to their applicability. Moreover, current research focuses on adapting converged models to previously unseen fabrics in a post hoc manner, rather than training models that generalize better in the first place. In our work, we explore this potential for the first time. Specifically, we propose that previously unseen fabrics can be regarded as shifts in the underlying data distribution. We therefore argue that factors which reportedly improve a model’s resistance to distribution shifts should also improve the performance of supervised fabric ad methods on unseen fabrics. Hence, we assess the potential benefits of: (I) vrm techniques adapted to the fabric ad use-case, (II) different loss functions, (III) ImageNet pre-training, (IV) dataset diversity, and (V) model architecture as well as model complexity. The subsequently performed large-scale analysis reveals that (I) only the vrm technique, AugMix, consistently improves performance on unseen fabrics; (II) hsc outperforms other loss functions when combined with AugMix and (III) ImageNet pre-training, which is already beneficial on its own; (IV) increasing dataset diversity improves performance on unseen fabrics; and (V) architectures with better ImageNet performance also perform better on unseen fabrics, yet the same does not hold for more complex models. Notably, the results show that not all factors and techniques which reportedly improve a model’s resistance to distribution shifts in natural images also improve the generalization of supervised fabric ad methods to unseen fabrics, demonstrating the necessity of our work. Additionally, we also assess whether the performance gains of models which generalize better propagate to post hoc adaptation methods and show this to be the case. Since no suitable fabric dataset was publicly available at the time of this work, we acquired our own fabric dataset, called OLP, as the basis for the above experiments. OLP consists of 38 complex, patterned fabrics, more than 6400 images in total, and is made publicly available.

show abstract

Estimating the Probability Density Function of New Fabrics for Fabric Anomaly Detection

Cited by 4 publications

References 0 publications

Gaussian Anomaly Detection by Modeling the Distribution of Normal Data in Pretrained Deep Features

Gaussian Anomaly Detection by Modeling the Distribution of Normal Data in Pretrained Deep Features

Transfer Learning Gaussian Anomaly Detection by Fine-tuning Representations

Increasing the Generalization of Supervised Fabric Anomaly Detection Methods to Unseen Fabrics

Contact Info

Product

Resources

About