Scalable Sketch-Based Image Retrieval Using Color Gradient Features

Tu Bui, John Collomosse; Proceedings of the IEEE International Conference on Computer Vision (ICCV) Workshops, 2015, pp. 1-8

Abstract


We present a scalable system for sketch-based image retrieval (SBIR), extending the state of the art Gradient Field HoG (GF-HoG) retrieval framework through two technical contributions. First, we extend GF-HoG to enable color-shape retrieval and comprehensively evaluate several early- and late-fusion approaches for integrating the modality of color, considering both the accuracy and speed of sketch retrieval. Second, we propose an efficient inverse-index representation for GF-HoG that delivers scalable search with interactive query times over millions of images. A mobile app demo accompanies this paper (Android) - see play.google.com/store/apps/details?id=com.collomosse.sketcher

Related Material


[pdf]
[bibtex]
@InProceedings{Bui_2015_ICCV_Workshops,
author = {Bui, Tu and Collomosse, John},
title = {Scalable Sketch-Based Image Retrieval Using Color Gradient Features},
booktitle = {Proceedings of the IEEE International Conference on Computer Vision (ICCV) Workshops},
month = {December},
year = {2015}
}