Learning Visual Similarity Measures for Comparing Never Seen Objects

@inproceedings{citeulike:1536147, abstract = {In this paper we propose and evaluate an algorithm that learns a similarity measure for comparing never seen objects. The measure is learned from pairs of training images labeled "same" or "different". This is far less informative than the commonly used individual image labels (e.g. "car model X"), but it is cheaper to obtain. The proposed algorithm learns the characteristic differences between local descriptors sampled from pairs of "same" and "different" images. These differences are vector quantized by an ensemble of extremely randomized binary trees, and the similarity measure is computed from the quantized differences. The extremely randomized trees are fast to learn, robust due to the redundant information they carry and they have been proved to be very good clusterers. Furthermore, the trees ef ciently combine different feature types (SIFT and geometry). We evaluate our innovative similarity measure on four very different datasets and consistantly outperform the state-of-the-art competitive approaches.}, author = {Nowak, Eric and Jurie, Frederic }, booktitle = {Computer Vision and Pattern Recognition, 2007. CVPR '07. IEEE Conference on}, citeulike-article-id = {1536147}, journal = {Computer Vision and Pattern Recognition, 2007. CVPR '07. IEEE Conference on}, keywords = {2007, category-recognition, cvpr, metric-learning, object-categorization, object-recognition}, pages = {1--8}, posted-at = {2007-08-05 09:25:55}, priority = {0}, title = {Learning Visual Similarity Measures for Comparing Never Seen Objects}, url = {http://ieeexplore.ieee.org/xpls/abs\_all.jsp?arnumber=4269994}, year = {2007} }

TY - CONF ID - citeulike:1536147 L3 - citeulike-article-id:1536147 TI - Learning Visual Similarity Measures for Comparing Never Seen Objects T2 - Computer Vision and Pattern Recognition, 2007. CVPR '07. IEEE Conference on JF - Computer Vision and Pattern Recognition, 2007. CVPR '07. IEEE Conference on SP - 1 EP - 8 N2 - In this paper we propose and evaluate an algorithm that learns a similarity measure for comparing never seen objects. The measure is learned from pairs of training images labeled "same" or "different". This is far less informative than the commonly used individual image labels (e.g. "car model X"), but it is cheaper to obtain. The proposed algorithm learns the characteristic differences between local descriptors sampled from pairs of "same" and "different" images. These differences are vector quantized by an ensemble of extremely randomized binary trees, and the similarity measure is computed from the quantized differences. The extremely randomized trees are fast to learn, robust due to the redundant information they carry and they have been proved to be very good clusterers. Furthermore, the trees ef ciently combine different feature types (SIFT and geometry). We evaluate our innovative similarity measure on four very different datasets and consistantly outperform the state-of-the-art competitive approaches. KW - 2007 KW - category-recognition KW - cvpr KW - metric-learning KW - object-categorization KW - object-recognition AU - Nowak, Eric AU - Jurie, Frederic PY - 2007/// UR - http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=4269994 ER -