-
Anomaly Detection via Multi-Scale Contrasted Memory
Authors:
Loic Jezequel,
Ngoc-Son Vu,
Jean Beaudet,
Aymeric Histace
Abstract:
Deep anomaly detection (AD) aims to provide robust and efficient classifiers for one-class and unbalanced settings. However current AD models still struggle on edge-case normal samples and are often unable to keep high performance over different scales of anomalies. Moreover, there currently does not exist a unified framework efficiently covering both one-class and unbalanced learnings. In the lig…
▽ More
Deep anomaly detection (AD) aims to provide robust and efficient classifiers for one-class and unbalanced settings. However current AD models still struggle on edge-case normal samples and are often unable to keep high performance over different scales of anomalies. Moreover, there currently does not exist a unified framework efficiently covering both one-class and unbalanced learnings. In the light of these limitations, we introduce a new two-stage anomaly detector which memorizes during training multi-scale normal prototypes to compute an anomaly deviation score. First, we simultaneously learn representations and memory modules on multiple scales using a novel memory-augmented contrastive learning. Then, we train an anomaly distance detector on the spatial deviation maps between prototypes and observations. Our model highly improves the state-of-the-art performance on a wide range of object, style and local anomalies with up to 50% error relative improvement on CIFAR-100. It is also the first model to keep high performance across the one-class and unbalanced settings.
△ Less
Submitted 9 March, 2023; v1 submitted 16 November, 2022;
originally announced November 2022.
-
Efficient Anomaly Detection Using Self-Supervised Multi-Cue Tasks
Authors:
Loic Jezequel,
Ngoc-Son Vu,
Jean Beaudet,
Aymeric Histace
Abstract:
Anomaly detection is important in many real-life applications. Recently, self-supervised learning has greatly helped deep anomaly detection by recognizing several geometric transformations. However these methods lack finer features, usually highly depend on the anomaly type, and do not perform well on fine-grained problems. To address these issues, we first introduce in this work three novel and e…
▽ More
Anomaly detection is important in many real-life applications. Recently, self-supervised learning has greatly helped deep anomaly detection by recognizing several geometric transformations. However these methods lack finer features, usually highly depend on the anomaly type, and do not perform well on fine-grained problems. To address these issues, we first introduce in this work three novel and efficient discriminative and generative tasks which have complementary strength: (i) a piece-wise jigsaw puzzle task focuses on structure cues; (ii) a tint rotation recognition is used within each piece, taking into account the colorimetry information; (iii) and a partial re-colorization task considers the image texture. In order to make the re-colorization task more object-oriented than background-oriented, we propose to include the contextual color information of the image border via an attention mechanism. We then present a new out-of-distribution detection function and highlight its better stability compared to existing methods. Along with it, we also experiment different score fusion functions. Finally, we evaluate our method on an extensive protocol composed of various anomaly types, from object anomalies, style anomalies with fine-grained classification to local anomalies with face anti-spoofing datasets. Our model significantly outperforms state-of-the-art with up to 36% relative error improvement on object anomalies and 40% on face anti-spoofing problems.
△ Less
Submitted 9 December, 2022; v1 submitted 24 November, 2021;
originally announced November 2021.
-
Fine-grained Anomaly Detection via Multi-task Self-Supervision
Authors:
Loic Jezequel,
Ngoc-Son Vu,
Jean Beaudet,
Aymeric Histace
Abstract:
Detecting anomalies using deep learning has become a major challenge over the last years, and is becoming increasingly promising in several fields. The introduction of self-supervised learning has greatly helped many methods including anomaly detection where simple geometric transformation recognition tasks are used. However these methods do not perform well on fine-grained problems since they lac…
▽ More
Detecting anomalies using deep learning has become a major challenge over the last years, and is becoming increasingly promising in several fields. The introduction of self-supervised learning has greatly helped many methods including anomaly detection where simple geometric transformation recognition tasks are used. However these methods do not perform well on fine-grained problems since they lack finer features. By combining in a multi-task framework high-scale shape features oriented task with low-scale fine features oriented task, our method greatly improves fine-grained anomaly detection. It outperforms state-of-the-art with up to 31% relative error reduction measured with AUROC on various anomaly detection problems.
△ Less
Submitted 17 March, 2022; v1 submitted 20 April, 2021;
originally announced April 2021.
-
Proceedings Cassting Workshop on Games for the Synthesis of Complex Systems and 3rd International Workshop on Synthesis of Complex Parameters
Authors:
Thomas Brihaye,
Benoît Delahaye,
Loïg Jezequel,
Nicolas Markey,
Jiří Srba
Abstract:
This volume contains the joint proceedings of the Workshop on Games for the Synthesis of Complex Systems (CASSTING'16) and of the 3rd International Workshop on Synthesis of Complex Parameters (SynCoP'16). The workshops were held in Eindhoven, The Netherlands, as satellite events of the 19th European Joint Conferences on Theory and Practice of Software (ETAPS'16). Both workshops are closely relate…
▽ More
This volume contains the joint proceedings of the Workshop on Games for the Synthesis of Complex Systems (CASSTING'16) and of the 3rd International Workshop on Synthesis of Complex Parameters (SynCoP'16). The workshops were held in Eindhoven, The Netherlands, as satellite events of the 19th European Joint Conferences on Theory and Practice of Software (ETAPS'16). Both workshops are closely related in their topics as well as target audience and they shared a joint invited talk given by Giorgio Delzanno.
△ Less
Submitted 30 July, 2016;
originally announced August 2016.
-
Computation of Summaries Using Net Unfoldings
Authors:
Javier Esparza,
Loïg Jezequel,
Stefan Schwoon
Abstract:
We study the following summarization problem: given a parallel composition A=A1||...||An of labelled transition systems communicating with the environment through a distinguished component Ai, efficiently compute a summary Si such that E||A and E||Si are trace-equivalent for every environment E. While Si can be computed using elementary automata theory, the resulting algorithm suffers from the sta…
▽ More
We study the following summarization problem: given a parallel composition A=A1||...||An of labelled transition systems communicating with the environment through a distinguished component Ai, efficiently compute a summary Si such that E||A and E||Si are trace-equivalent for every environment E. While Si can be computed using elementary automata theory, the resulting algorithm suffers from the state-explosion problem. We present a new, simple but subtle algorithm based on net unfoldings, a partial-order semantics, give some experimental results using an implementation on top of MOLE, and show that our algorithm can handle divergences and compute weighted summaries with minor modifications.
△ Less
Submitted 8 October, 2013;
originally announced October 2013.
-
Méthode de calcul du rayonnement acoustique de structures complexes
Authors:
Marianne Viallet,
Gérald Poumérol,
Olivier Dessombz,
Louis Jezequel
Abstract:
In the automotive industry, predicting noise during design cycle is a necessary step. Well-known methods exist to answer this issue in low frequency domain. Among these, Finite Element Methods, adapted to closed domains, are quite easy to implement whereas Boundary Element Methods are more adapted to infinite domains, but may induce singularity problems. In this article, the described method, th…
▽ More
In the automotive industry, predicting noise during design cycle is a necessary step. Well-known methods exist to answer this issue in low frequency domain. Among these, Finite Element Methods, adapted to closed domains, are quite easy to implement whereas Boundary Element Methods are more adapted to infinite domains, but may induce singularity problems. In this article, the described method, the SDM, allows to use both methods in their best application domain. A new method is also presented to solve the SDM exterior problem. Instead of using Boundary Element Methods, an original use of Finite Elements is made. Efficiency of this new version of the Substructure Deletion Method is discussed.
△ Less
Submitted 8 April, 2008;
originally announced April 2008.
