Decomposed learning for joint object segmentation and categorization

Yi Hsuan Tsai, Jimei Yang, Ming Hsuan Yang

Research output: Contribution to conferencePaper

3 Citations (Scopus)

Abstract

We present a learning algorithm for joint object segmentation and categorization that decomposes the original problem into two sub-tasks and admits their bidirectional interaction. In the first stage, in order to decompose output space, we train category-specific segmentation models to generate figure-ground hypotheses. In the second stage, by taking advantage of object figure-ground information, we train a multi-class segment-based categorization model to determine the object class. A re-ranking strategy is then applied to classified segments to obtain the final category-level segmentation results. Experiments on the Graz-02 and Caltech-101 datasets show that the proposed algorithm performs favorably against the state-of-the-art methods.

Original languageEnglish
DOIs
Publication statusPublished - 2013 Jan 1
Event2013 24th British Machine Vision Conference, BMVC 2013 - Bristol, United Kingdom
Duration: 2013 Sep 92013 Sep 13

Conference

Conference2013 24th British Machine Vision Conference, BMVC 2013
CountryUnited Kingdom
CityBristol
Period13/9/913/9/13

Fingerprint

Learning algorithms
Experiments

All Science Journal Classification (ASJC) codes

  • Computer Vision and Pattern Recognition

Cite this

Tsai, Y. H., Yang, J., & Yang, M. H. (2013). Decomposed learning for joint object segmentation and categorization. Paper presented at 2013 24th British Machine Vision Conference, BMVC 2013, Bristol, United Kingdom. https://doi.org/10.5244/C.27.80
Tsai, Yi Hsuan ; Yang, Jimei ; Yang, Ming Hsuan. / Decomposed learning for joint object segmentation and categorization. Paper presented at 2013 24th British Machine Vision Conference, BMVC 2013, Bristol, United Kingdom.
@conference{23222e8f5def404c94236dfd9b2465a2,
title = "Decomposed learning for joint object segmentation and categorization",
abstract = "We present a learning algorithm for joint object segmentation and categorization that decomposes the original problem into two sub-tasks and admits their bidirectional interaction. In the first stage, in order to decompose output space, we train category-specific segmentation models to generate figure-ground hypotheses. In the second stage, by taking advantage of object figure-ground information, we train a multi-class segment-based categorization model to determine the object class. A re-ranking strategy is then applied to classified segments to obtain the final category-level segmentation results. Experiments on the Graz-02 and Caltech-101 datasets show that the proposed algorithm performs favorably against the state-of-the-art methods.",
author = "Tsai, {Yi Hsuan} and Jimei Yang and Yang, {Ming Hsuan}",
year = "2013",
month = "1",
day = "1",
doi = "10.5244/C.27.80",
language = "English",
note = "2013 24th British Machine Vision Conference, BMVC 2013 ; Conference date: 09-09-2013 Through 13-09-2013",

}

Tsai, YH, Yang, J & Yang, MH 2013, 'Decomposed learning for joint object segmentation and categorization', Paper presented at 2013 24th British Machine Vision Conference, BMVC 2013, Bristol, United Kingdom, 13/9/9 - 13/9/13. https://doi.org/10.5244/C.27.80

Decomposed learning for joint object segmentation and categorization. / Tsai, Yi Hsuan; Yang, Jimei; Yang, Ming Hsuan.

2013. Paper presented at 2013 24th British Machine Vision Conference, BMVC 2013, Bristol, United Kingdom.

Research output: Contribution to conferencePaper

TY - CONF

T1 - Decomposed learning for joint object segmentation and categorization

AU - Tsai, Yi Hsuan

AU - Yang, Jimei

AU - Yang, Ming Hsuan

PY - 2013/1/1

Y1 - 2013/1/1

N2 - We present a learning algorithm for joint object segmentation and categorization that decomposes the original problem into two sub-tasks and admits their bidirectional interaction. In the first stage, in order to decompose output space, we train category-specific segmentation models to generate figure-ground hypotheses. In the second stage, by taking advantage of object figure-ground information, we train a multi-class segment-based categorization model to determine the object class. A re-ranking strategy is then applied to classified segments to obtain the final category-level segmentation results. Experiments on the Graz-02 and Caltech-101 datasets show that the proposed algorithm performs favorably against the state-of-the-art methods.

AB - We present a learning algorithm for joint object segmentation and categorization that decomposes the original problem into two sub-tasks and admits their bidirectional interaction. In the first stage, in order to decompose output space, we train category-specific segmentation models to generate figure-ground hypotheses. In the second stage, by taking advantage of object figure-ground information, we train a multi-class segment-based categorization model to determine the object class. A re-ranking strategy is then applied to classified segments to obtain the final category-level segmentation results. Experiments on the Graz-02 and Caltech-101 datasets show that the proposed algorithm performs favorably against the state-of-the-art methods.

UR - http://www.scopus.com/inward/record.url?scp=84898431787&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84898431787&partnerID=8YFLogxK

U2 - 10.5244/C.27.80

DO - 10.5244/C.27.80

M3 - Paper

AN - SCOPUS:84898431787

ER -

Tsai YH, Yang J, Yang MH. Decomposed learning for joint object segmentation and categorization. 2013. Paper presented at 2013 24th British Machine Vision Conference, BMVC 2013, Bristol, United Kingdom. https://doi.org/10.5244/C.27.80