Abstract

Scene classification is the task of determining the scene type in which a photograph was taken. In this paper we present a novel local descriptor suited for such a task: Oriented Texture Curves (OTC). Our descriptor captures the texture of a patch along multiple orientations, while maintaining robustness to illumination changes, geometric distortions and local contrast differences. We show that our descriptor outperforms all state-of-the-art descriptors for scene classification algorithms on the most extensive scene classification benchmark to-date.

Paper

Ran Margolin, Lihi Zelnik-Manor and Ayellet Tal: “OTC: A Novel Local Descriptor for Scene Classifcation”, ECCV 2014 [pdf]

Results

[1]	Lazebnik et al. CVPR'06	[4]	Xiao et al. CVPR'10	[7]	Su et al. IJCV'12
[2]	Shechtman et al. CVPR'07	[5]	Felzenszwalb etl al. PAMI'10	[8]	Donahue et al. ICML'14
[3]	Shen et al. CVPR'13	[6]	Kwitt et al. ECCV'12	[9]	Gong et al. CoRR'14

Code

OTC descriptor for a single image [download] - Tested on Windows 64-bit Matlab 2012b & 2013a.
The code contains the Matlab function im2colstep.m, written by Dr. Ron Rubinstein.

The code is for academic purposes only. Please cite this paper if you make use of it:

@conference{margoinOTC, title={OTC: A Novel Local Descriptor for Scene Classification}, author={Margolin, R. and Zelnik-Manor, L. and Tal, A}, year = {2014}, booktitle = {ECCV}}

In case of any problems please contact us at: margolin (at) tx.technion.ac.il or hovav (at) ee.technion.ac.il