Discriminative Mixture-of-Templates for Viewpoint Classification

DMCA

Discriminative Mixture-of-Templates for Viewpoint Classification (2010)

Cached

Download Links

by Chunhui Gu , Xiaofeng Ren

Venue:	EUROPEAN CONFERENCE ON COMPUTER VISION
Citations:	59 - 3 self

BibTeX

@INPROCEEDINGS{Gu10discriminativemixture-of-templates,
    author = {Chunhui Gu and Xiaofeng Ren},
    title = {Discriminative Mixture-of-Templates for Viewpoint Classification},
    booktitle = {EUROPEAN CONFERENCE ON COMPUTER VISION},
    year = {2010},
    publisher = {}
}

OpenURL

Abstract

Object viewpoint classification aims at predicting an approximate 3D pose of objects in a scene and is receiving increasing attention. State-of-the-art approaches to viewpoint classification use generative models to capture relations between object parts. In this work we propose to use a mixture of holistic templates (e.g. HOG) and discriminative learning for joint viewpoint classification and category detection. Inspired by the work of Felzenszwalb et al 2009, we discriminatively train multiple components simultaneously for each object category. A large number of components are learned in the mixture and they are associated with canonical viewpoints of the object through different levels of supervision, being fully supervised, semi-supervised, or unsupervised. We show that discriminative learning is capable of producing mixture components that directly provide robust viewpoint classification, significantly outperforming the state of the art: we improve the viewpoint accuracy on the Savarese et al 3D Object database from 57% to 74%, and that on the VOC 2006 car database from 73% to 86%. In addition, the mixture-of-templates approach to object viewpoint/pose has a natural extension to the continuous case by discriminatively learning a linear appearance model locally at each discrete view. We evaluate continuous viewpoint estimation on a dataset of everyday objects collected using IMUs for groundtruth annotation: our mixture model shows great promise comparing to a number of baselines including discrete nearest neighbor and linear regression.

Keyphrases

discriminative mixture-of-templates viewpoint classification discriminative learning object category mixture-of-templates approach state-of-the-art approach large number generative model object database everyday object joint viewpoint classification holistic template savarese et category detection mixture model robust viewpoint classification different level linear appearance model object part continuous case multiple component viewpoint pose continuous viewpoint estimation canonical viewpoint discrete view object viewpoint classification car database viewpoint accuracy linear regression mixture component natural extension great promise groundtruth annotation

DMCA