Academic
Publications
Efficient Learning of Relational Object Class Models

Efficient Learning of Relational Object Class Models,10.1007/s11263-007-0091-7,International Journal of Computer Vision,Aharon Bar-hillel,Daphna Weins

Efficient Learning of Relational Object Class Models   (Citations: 12)
BibTex | RIS | RefWorks Download
We present an efficient method for learning part-based object class models. The models include location and scale relations between parts, as well as part appearance. Mod- els are learnt from raw object and background images, rep- resented as an unordered set of features extracted using an interest point detector. The object class is generatively mod- eled using a simple Bayesian network with a central hidden node containing location and scale information, and nodes describing object parts. The model's parameters, however, are optimized to reduce a loss function which reflects train- ing error, as in discriminative methods. Specifically, the optimization is done using a boosting-like technique with complexity linear in the number of parts and the number of features per image. This efficiency allows our method to learn relational models with many parts and features, and leads to improved results when compared with other methods. Extensive experimental results are described, us- ing some common bench-mark datasets and three sets of newly collected data, showing the relative advantage of our method. In our current work, we try to enjoy the benefits of both worlds: The modeling power of the generative approach, with the accuracy and efficiency of discriminative optimiza- tion. We present a novel method for object class recogni- tion, based on discriminative optimization of a simple gen- erative object model. Specifically, we use a compact star- like Bayesian network as our generative model, and extend current discriminative boosting techniques to enable param-
Journal: International Journal of Computer Vision - IJCV , vol. 77, no. 1-3, pp. 175-198, 2008
Cumulative Annual
View Publication
The following links allow you to view full publications. These links are maintained by other sources not affiliated with Microsoft Academic Search.
    • ...It has been more difficult to train partbased models discriminatively, though strategies exist [4], [23], [32], [34]...

    Pedro F. Felzenszwalbet al. Object Detection with Discriminatively Trained Part-Based Models

    • ...Part-based representations have attracted a lot of machine vision research [14,19,4,20,21], and are believed to play an important role in human vision [22]...
    • ...Part localization adds a score to the feature reflecting the location of the part in an absolute framework (commonly referred to as a ’star model’ [21]), or with respect to other parts (e.g...
    • ...of the detection location l. Such features represent location sensitive part detection, attaining a high value when both the appearance score is high and the position is close to the Gaussian mean, similar to parts in a star-like model [21]...

    Aharon Bar-Hillelet al. Part-Based Feature Synthesis for Human Detection

    • ...and D 2 [0;1] measures the difference in orientations, Dl 2 [0;1] the difference in lengths, and Dv 2 [0;2] the difference in relative primitive positions...
    • ...It is easily verified that a codeword which finds a token that satisfies its appearance matching constraint and that is in its estimated window has<( ) value in the range [0; 2], where a lower value indicates better matching...
    • ...<i(^s;^ i and <j(^s;^ j ; where is a threshold in the range [0; 2]. Note that since this xCC is matched at the hypothesis, it therefore implicitly models the appearance and geometric configurations of shape tokens at this hypothesis...
    • ...Object category Our method Sivic ’05 [20] Crandall ’06 [8] Fergus ’07 [9] Bar-Hillel ’08 [2] Chen ’09 [5] Zhu ’09 [21]...

    Alex Y. S. Chiaet al. Object recognition by discriminative combinations of line segments and...

    • ...According to the psychophysics of human saliency, humans can easily distinguish differences between targets and distracting background along a single feature channel (e.g., different color), but not when they require a connection of features (e.g., differences in the connection of color and shape) [19, 20, 21, 22]...
    • ...If a feature is present in the region of target (Y=1), but absent in the distracting background (Y=0), then it passes the test as follows [19, 21, 22]...

    Bing Wanget al. Saliency distinguishing and applications to semantics extraction and r...

    • ...The BoF methods typically utilise interest point detectors and descriptors instead of fixed local parts, but several particularly successful hybrids of interest points and constellation search have been proposed [1, 2]...

    Joni-Kristian Kamarainenet al. Learning and Detection of Object Landmarks in Canonical Object Space

Sort by: