For Full-Text PDF, please login, if you are a member of IEICE,|
or go to Pay Per View on menu list, if you are a nonmember of IEICE.
Specific Person Image Retrieval from Surveillance Videos via Condition-Separating Relevance Feedback
Yohei ISEKI Yasutomo KAWANISHI Masayuki MUKUNOKI Michihiko MINOH
D - Abstracts of IEICE TRANSACTIONS on Information and Systems (Japanese Edition)
Publication Date: 2015/01/01
Online ISSN: 1881-0225
Type of Manuscript: PAPER
image retrieval, person re-identification, Specific Person Retrieaval System, relevance feedback,
Full Text(in Japanese): PDF(1.4MB)
>>Buy this Article
We present a novel method of Relevance Feedback for Specific Person Image Retrieval: Condition-Separating Relevance Feedback (CSRF). Image features of a person vary according to their observation conditions. When image features of a wanted person on an observation condition are similar to image features of some persons on different observation conditions, their images are ranked higher. Handling user's feedbacks for images on each observation condition separately, CSRF makes their images rank lower and pulls up the recall of the retrieval. Evaluation result shows CSRF gets the best score on public datasets.