A Unified Approach for Detection, Classification and Segmentation - Robotics Institute Carnegie Mellon University

A Unified Approach for Detection, Classification and Segmentation

Derek Hoiem, Santosh Kumar Divvala, James H. Hays, Alexei A. Efros, and Martial Hebert
Workshop Paper, ECCV '08 The PASCAL Visual Object Classes Challenge Workshop, October, 2008

Abstract

To tackle the challenging dataset presented in PASCAL VOC 2008 challenge, we use a highly successful appearance-based detector and augment it with rich contextual cues extracted from the image to further improve its performance. Specifically, we train detectors to obtain the confidence that a window contains an object based solely on global scene statistics, nearby regions, the object position and size, geographic context and boundaries. Our interest is to study how much each of these contextual cues can add to the performance of the local appearance based detector.

BibTeX

@workshop{Hoiem-2008-10118,
author = {Derek Hoiem and Santosh Kumar Divvala and James H. Hays and Alexei A. Efros and Martial Hebert},
title = {A Unified Approach for Detection, Classification and Segmentation},
booktitle = {Proceedings of ECCV '08 The PASCAL Visual Object Classes Challenge Workshop},
year = {2008},
month = {October},
}