Abnormal Object Detection by Canonical Scene-Based Contextual Model

Abstract

Abstract. Contextual modeling is a critical issue in scene understanding. Object detection accuracy can be improved by exploiting tendencies that are common among object configurations. However, conventional contextual models only exploit the tendencies of normal objects; abnormal objects that do not follow the same tendencies are hard to detect through contextual model. This paper proposes a novel generative model that detects abnormal objects by meeting four proposed criteria of success. This model generates normal as well as abnormal objects, each following their respective tendencies. Moreover, this generation is controlled by a latent scene variable. All latent variables of the proposed model are predicted through optimization via population-based Markov Chain Monte Carlo, which has a relatively short convergence time. We present a new abnormal dataset classified into three categories to thoroughly measure the accuracy of the proposed model for each category; the results demonstrate the superiority of our proposed approach over existing methods.

Object-Centric Anomaly Detection by Attribute-Based Reasoning

“…However, in this paper we are mainly focused on abnormalities stemming from the object itself regardless of the context. In that sense our work is complementary to [22,24].…”

“…Very recently, out-of-context objects have been studied in [22,24]. [22] uses a latent support graph to model the context and [24] use a generative model that learns for multiple criteria of normality and abnormality.…”

“…[22] uses a latent support graph to model the context and [24] use a generative model that learns for multiple criteria of normality and abnormality. Contextual irregularities are one of the reasons of abnormality.…”

Local visual pattern modelling for image and video classification

“…The work [80] focuses on finding abnormal objects in given scenes. They consider wider range of irregular objects like those violate co-occurrence with surrounding objects or violate expected scale.…”

“…Mimicking this ability with computer vision technique can be practically useful for the applications such as surveillance or quality control. Existing studies towards this goal are usually conducted on small datasets and controlled scenarios i.e., with relatively simple background [7] or specific type of unusuality [15,80]. To address this issue, in this work we present a large dataset which captures more general unusualities and has more complex background.…”