TY - JOUR
T1 - Local features and kernels for classification of texture and object categories
T2 - A comprehensive study
AU - Zhang, J.
AU - Marszałek, M.
AU - Lazebnik, S.
AU - Schmid, C.
N1 - Funding Information:
This research was supported by the French project Mo-ViStaR under the program “ACI Masse de données”, the European project LAVA, the European Network of Excellence PASCAL, and the UIUC-CNRS-INRIA collaboration agreement. J. Zhang was funded by an ACI postdoctoral fellowship and M. Marszalek by the INRIA student exchange program and a grant from the European Community under the Marie-Curie project VISITOR. S. Lazebnik was funded in part by Toyota and National Science Foundation grants IIS-0308087 and IIS-0312438. We also thank J. Ponce for discussions, M. Varma and A. Zisserman for providing the subset of the CUReT dataset used in their paper, E. Hayman for explaining the implementation details of their method, and Y. Rubner for making his implementation of EMD publicly available on the web.
PY - 2007/6
Y1 - 2007/6
N2 - Recently, methods based on local image features have shown promise for texture and object recognition tasks. This paper presents a large-scale evaluation of an approach that represents images as distributions (signatures or histograms) of features extracted from a sparse set of keypoint locations and learns a Support Vector Machine classifier with kernels based on two effective measures for comparing distributions, the Earth Mover's Distance and the χ2 distance. We first evaluate the performance of our approach with different keypoint detectors and descriptors, as well as different kernels and classifiers. We then conduct a comparative evaluation with several state-of-the-art recognition methods on four texture and five object databases. On most of these databases, our implementation exceeds the best reported results and achieves comparable performance on the rest. Finally, we investigate the influence of background correlations on recognition performance via extensive tests on the PASCAL database, for which ground-truth object localization information is available. Our experiments demonstrate that image representations based on distributions of local features are surprisingly effective for classification of texture and object images under challenging real-world conditions, including significant intra-class variations and substantial background clutter.
AB - Recently, methods based on local image features have shown promise for texture and object recognition tasks. This paper presents a large-scale evaluation of an approach that represents images as distributions (signatures or histograms) of features extracted from a sparse set of keypoint locations and learns a Support Vector Machine classifier with kernels based on two effective measures for comparing distributions, the Earth Mover's Distance and the χ2 distance. We first evaluate the performance of our approach with different keypoint detectors and descriptors, as well as different kernels and classifiers. We then conduct a comparative evaluation with several state-of-the-art recognition methods on four texture and five object databases. On most of these databases, our implementation exceeds the best reported results and achieves comparable performance on the rest. Finally, we investigate the influence of background correlations on recognition performance via extensive tests on the PASCAL database, for which ground-truth object localization information is available. Our experiments demonstrate that image representations based on distributions of local features are surprisingly effective for classification of texture and object images under challenging real-world conditions, including significant intra-class variations and substantial background clutter.
KW - Image classification
KW - Kernel methods
KW - Object recognition
KW - Scale- and affine-invariant keypoints
KW - Support vector machines
KW - Texture recognition
UR - http://www.scopus.com/inward/record.url?scp=33846580425&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=33846580425&partnerID=8YFLogxK
U2 - 10.1007/s11263-006-9794-4
DO - 10.1007/s11263-006-9794-4
M3 - Article
AN - SCOPUS:33846580425
SN - 0920-5691
VL - 73
SP - 213
EP - 238
JO - International Journal of Computer Vision
JF - International Journal of Computer Vision
IS - 2
ER -