Object-based Image Classification and Object Detection
Shuai Zheng1 Junge Zhang1 Yinan Yu1 Yongzhen Huang1 Weiqiang Ren1, Kaiqi Huang1, Tieniu Tan1
1NLPR, Institute of Automation, Chinese Academy of Sciences
Built on the top of deformable part based model, an very exhausted feature learning and ensemble learning research has been conducted to push part-based method to the performance limits.
PASCAL VOC Challenge provides a platform to evaluate how object based image classification and object detection work on the real image data(e.g. Flickr images). The challenge database contains 20 classes of object. All the image are collected from Flickr public available images. The purpose of image classification is to identify the class label of the objects in given an image. Based on the existing state-of-the-art Bag-of-words framework, we improved it by developing three new components including saliency coding, code relation modeling, and dictionary learning. Insipired by the existing state-of-the-art deformable part based model framework, we develop an efficient but complicated approaches to achieve both low-level feature combinations and high-level hybrid approach combinations. Recently, some new improvements leads NLPR to win the two tasks object detection and image classification at PASCAL VOC Challenge 2011 again.Find more information at PASCAL VOC Challenge website. link.
Overview diagram of the object detection approach
Overview Summary of object based image classification approach
Results of Detection at VOC 2010
Publications and Presentations
Y. Yu, J. Zhang, Y. Huang, S. Zheng, W. Ren, K. Huang, and T. Tan, “Object Detection by Context and Boosted HOG-LBP“, in PASCAL Visual Object Challenge Workshop, European Conference on Computer Vision (ECCV), Crete, Greece, 2010.(No.1) [bib]
Y. Huang, S. Zheng, W. Ren, Y. Yu, J. Zhang, K. Huang, and T. Tan, “SVM classifier and saliency coding for Image classification“, in PASCAL Visual Object Challenge Workshop, European Conference on Computer Vision (ECCV), Crete, Greece, 2010.
J. Zhang, Y. Yu, S. Zheng, K. Huang, and T. Tan, “An Empirical Study of Visual Features for Part Based Model,” in Asian Conference on Pattern Recognition (ACPR), Beijing, China, 2011. [bib]
Efficient Implementation of Bag of words [code]