Fast concurrent object classification and localization
Object localization and classification are important problems incomputer vision. However, in many applications, exhaustive searchover all class labels and image locations is computationallyprohibitive. While several methods have been proposed to makeeither classification or localization more efficie...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Published: |
2008
|
| Online Access: | http://hdl.handle.net/1721.1/41862 |
| _version_ | 1826194161140760576 |
|---|---|
| author | Yeh, Tom Lee, John J. Darrell, Trevor |
| author2 | Trevor Darrell |
| author_facet | Trevor Darrell Yeh, Tom Lee, John J. Darrell, Trevor |
| author_sort | Yeh, Tom |
| collection | MIT |
| description | Object localization and classification are important problems incomputer vision. However, in many applications, exhaustive searchover all class labels and image locations is computationallyprohibitive. While several methods have been proposed to makeeither classification or localization more efficient, few havedealt with both tasks simultaneously. This paper proposes anefficient method for concurrent object localization andclassification based on a data-dependent multi-classbranch-and-bound formalism. Existing bag-of-featuresclassification schemes, which can be expressed as weightedcombinations of feature counts can be readily adapted to ourmethod. We present experimental results that demonstrate the meritof our algorithm in terms of classification accuracy, localizationaccuracy, and speed, compared to baseline approaches includingexhaustive search, the ISM method, and single-class branch andbound. |
| first_indexed | 2024-09-23T09:51:52Z |
| id | mit-1721.1/41862 |
| institution | Massachusetts Institute of Technology |
| last_indexed | 2024-09-23T09:51:52Z |
| publishDate | 2008 |
| record_format | dspace |
| spelling | mit-1721.1/418622019-04-10T17:42:01Z Fast concurrent object classification and localization Yeh, Tom Lee, John J. Darrell, Trevor Trevor Darrell Vision Object localization and classification are important problems incomputer vision. However, in many applications, exhaustive searchover all class labels and image locations is computationallyprohibitive. While several methods have been proposed to makeeither classification or localization more efficient, few havedealt with both tasks simultaneously. This paper proposes anefficient method for concurrent object localization andclassification based on a data-dependent multi-classbranch-and-bound formalism. Existing bag-of-featuresclassification schemes, which can be expressed as weightedcombinations of feature counts can be readily adapted to ourmethod. We present experimental results that demonstrate the meritof our algorithm in terms of classification accuracy, localizationaccuracy, and speed, compared to baseline approaches includingexhaustive search, the ISM method, and single-class branch andbound. 2008-06-11T20:15:30Z 2008-06-11T20:15:30Z 2008-06-10 MIT-CSAIL-TR-2008-033 http://hdl.handle.net/1721.1/41862 Massachusetts Institute of Technology Computer Science and Artificial Intelligence Laboratory 9 p. application/pdf application/postscript |
| spellingShingle | Yeh, Tom Lee, John J. Darrell, Trevor Fast concurrent object classification and localization |
| title | Fast concurrent object classification and localization |
| title_full | Fast concurrent object classification and localization |
| title_fullStr | Fast concurrent object classification and localization |
| title_full_unstemmed | Fast concurrent object classification and localization |
| title_short | Fast concurrent object classification and localization |
| title_sort | fast concurrent object classification and localization |
| url | http://hdl.handle.net/1721.1/41862 |
| work_keys_str_mv | AT yehtom fastconcurrentobjectclassificationandlocalization AT leejohnj fastconcurrentobjectclassificationandlocalization AT darrelltrevor fastconcurrentobjectclassificationandlocalization |