Deep Anomaly Detection Using Geometric Transformations

By Izhak Golan, Ran El-Yaniv. 2018

675 citations recorded by the source corpus.

We consider the problem of anomaly detection in images, and present a new detection technique. Given a sample of images, all known to belong to a "normal" class (e.g., dogs), we show how to train a deep neural model that can detect out-of-distribution images (i.e., non-dog objects). The main idea behind our scheme is to train a multi-class model to discriminate between dozens of geometric transformations applied on all the given images. The auxiliary expertise learned by the model generates feature detectors that effectively identify, at test time, anomalous images based on the softmax activation statistics of the model when applied on transformed images. We present extensive experiments using the proposed detector, which indicate that our algorithm improves state-of-the-art methods by a wide margin.

Subfields: Artificial Intelligence, Computer Networks and Communications, Signal Processing

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

In Computer Networks and Communications

In Signal Processing

Citations in the corpus

Out-of-Domain Detection for Natural Language Understanding in Dialog Systems (2019, methodology): “[27], using classifier ensembles [13], [28], applying Bayesian models [29], relying on distances to nearest-neighbors [9], [30], or even explicitly learning a detection score [31].”
Anomalous Instance Detection in Deep Learning: A Survey (2020, methodology): “Unsupervised [29] Learning to discriminate between geometric transformations is used for learning unique features that are useful in OOD detection”
Anomaly Detection via Reverse Distillation from One-Class Embedding (2022, methodology): “Category/Method MKD [33] Ours GT [10] GN [2] US [4] PSVDD [43] DAAD [16] MF [40] PaDiM [8] CutPaste [23] Ours”
Deep Learning for Anomaly Detection (2020, methodology): “, training data contains normal instances only, is assumed in [48].”
Continual Learning Based on OOD Detection and Task Masking (2022, methodology): “The OOD detection algorithm in CLOM is inspired by several recent advances in self-supervised learning, data augmentation [21], contrastive learning [11, 48], and their applications to OOD detection [18, 23, 65].”
Sub-Image Anomaly Detection with Deep Pyramid Correspondences (2020, background): “Golan and El-Yaniv [11] proposed a RotNet-based [10] approach, which performs geometric transformations on the input data and trains a network that attempts to recognize the transformation used.”
Towards Total Recall in Industrial Anomaly Detection (2021, methodology): “Other unsupervised representation learning methods can similarly be utilised, such as via GANs [13], learning to predict predeﬁned geometric transformations [20] or via normalizing ﬂows [42].”
Multiresolution Knowledge Distillation for Anomaly Detection (2020, methodology): “DSVDD[33] Auto-Encoder[55] AnoGan[41] VAE-GAN[5] Pix2Pix[20] GANomaly[3] Cycle-GAN[57] P-Net[56] GT[18] OURS”
A Unified Model for Multi-class Anomaly Detection (2022, methodology): “CNN is designed based on the ResNet-34 [7] by revising the followings.”
HRN: A Holistic Approach to One Class Learning (2020, background): “They almost exclusively use the neural network representation learning capability to generate a latent representation of the given class [1, 14, 16, 25, 51, 65, 68, 71, 83, 92, 95, 69, 73, 26, 82].”
Lips Don't Lie: A Generalisable and Robust Approach to Face Forgery Detection (2020, result): “Such pretraining is consistent with recent anomaly detection literature, which suggests that training on the “normal” class (in this case real videos) for a suitable task promotes learning of features that are useful for detecting “anomalous” samples (in this case fake videos) [23, 7].”
Memorizing Normality to Detect Anomaly: Memory-Augmented Deep Autoencoder for Unsupervised Anomaly Detection (2019, background): “The unsupervised anomaly detection [47, 43, 48, 32, 7] is to learn a normal profile given only the normal data examples and then identify the samples not conforming to the normal profile as anomalies, which is challenging due to E n c o d e r”
Self-Supervised Representation Learning: Introduction, advances, and challenges (2021, neutral): “For example, current state-of-the-art anomaly detectors [77] rely on SSRL training of rotation prediction, with the rotation prediction accuracy providing the anomaly score.”
Deep Semi-Supervised Anomaly Detection (2019, methodology): “The results in section 4.2 and other recent works [56, 22, 25] demonstrate that deep methods are especially superior on complex data with hierarchical structure.”
Detecting semantic anomalies (2019, methodology): “We discourage the recently-adopted practice of treating one category as in-distribution and many other categories as out-distributions (as in Pidhorskyi, Almohsen, and Doretto (2018) and Golan and El-Yaniv (2018), for example).”
Same Same But DifferNet: Semi-Supervised Defect Detection with Normalizing Flows (2020, methodology): “GeoTrans computes an anomaly score based on the classification of conducted transformations.”