Object Recognition Based on Efficient Sub-window Search

Authors:
Qing Nie;Shouyi Zhan;Weiming Li
Affiliations:
School of Information and Electronic, Beijing Institute of Technology, Beijing, China 100081;School of Computer Science, Beijing Institute of Technology, Beijing, China 100081;School of Information and Electronic, Beijing Institute of Technology, Beijing, China 100081
Venue:
AICI '09 Proceedings of the International Conference on Artificial Intelligence and Computational Intelligence
Year:
2009

Citing 6
Cited 0

Spatial Priors for Part-Based Recognition Using Statistical Models

CVPR '05 Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Volume 1 - Volume 01
Histograms of Oriented Gradients for Human Detection

CVPR '05 Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Volume 1 - Volume 01
Contour-Based Learning for Object Detection

ICCV '05 Proceedings of the Tenth IEEE International Conference on Computer Vision (ICCV'05) Volume 1 - Volume 01
Extremely randomized trees

Machine Learning
Groups of Adjacent Contour Segments for Object Detection

IEEE Transactions on Pattern Analysis and Machine Intelligence
A boundary-fragment-model for object detection

ECCV'06 Proceedings of the 9th European conference on Computer Vision - Volume Part II

Quantified Score

Hi-index	0.00

Visualization

Abstract

We propose a new method for object recognition in natural images. This method integrates bag of features model with efficient sub-window search technology. sPACT is introduces as local feature descriptor for recognition task. It can capture both local structures and global structures of an image patch efficiently by histogram of Census Transform. An efficient sub-window search method is adapted to perform localization. This method relies on a branch-and-bound scheme to find the global optimum of the quality function over all possible sub-windows. It requires much fewer classifier evaluations than the usually way does. The evaluation on PASCAL 2007 VOC dataset shows that this object recognition method has many advantages. It uses weakly supervised training method, yet has comparable localization performance to state-of-the-art algorithms. The feature descriptor can efficiently encode image patches, and localization method is fast without losing precision.