Object class detection

Abstract

Object class detection, also known as category-level object detection, has become one of the most focused areas in computer vision in the new century. This article attempts to provide a comprehensive survey of the recent technical achievements in this area of research. More than 270 major publications are included in this survey covering different aspects of the research, which include: (i) problem description: key tasks and challenges; (ii) core techniques: appearance modeling, localization strategies, and supervised classification methods; (iii) evaluation issues: approaches, metrics, standard datasets, and state-of-the-art results; and (iv) new development: particularly new approaches and applications motivated by the recent boom of social images. Finally, in retrospect of what has been achieved so far, the survey also discusses what the future may hold for object class detection research.

Review of Local Descriptor in RGB-D Object Recognition

“…While the challenges belonging to the second group include very large objects of different categories, high-dimensional descriptors and difficulties in obtaining labelled training samples without any ambiguity etc. [3].…”

“…To address these two challenges, [3] argues that there are three aspects involved, namely modelling appearance, localization strategies and supervised classification. The focus of the researchers was trying to develop techniques and algorithms in those three aspects in order to improve the visual object recognition system performance.…”

“…The focus of the researchers was trying to develop techniques and algorithms in those three aspects in order to improve the visual object recognition system performance. Among these three aspects, modelling appearance is the most important aspect [3]. Appearance modelling is focused on the selection of features that can handle various types of intra-class variations and can capture the discriminative aspects of the different categories.…”

“…While the challenges belonging to the second group include very large objects of different categories, high-dimensional descriptors and difficulties in obtaining labelled training samples without any ambiguity etc. [3].To address these two challenges, [3] argues that there are three aspects involved, namely modelling appearance, localization strategies and supervised classification. The focus of the researchers was trying to develop techniques and algorithms in those three aspects in order to improve the visual object recognition system performance.…”

A Performance Evaluation of Content-Based Image Retrieval for Scene Categorization

“…Thus, this can be used to categorize the scene accordingly. Object recognition in an image is the fundamental challenge [6] in computer vision. In this work, we interested in recognizing scene categories using the image taken by normal rectilinear camera lens.…”