Academic
Publications
Principled Detection-by-Classification From Multiple Views

Principled Detection-by-Classification From Multiple Views,Jérôme Berclaz,François Fleuret,Pascal Fua

Principled Detection-by-Classification From Multiple Views   (Citations: 6)
BibTex | RIS | RefWorks Download
Machine-learning based classification techniques have been shown to be effective at detecting objects in com- plex scenes. However, the final results are often obtained from the alarms produced by the classifiers through a post-processing which typically relies on ad hoc heuristics. Spatially close alarms are assumed to be triggered by the same target and grouped together. Here we replace those heuristics by a principled Bayesian approach, which uses knowledge about both the classifier response model and the scene geometry to combine multiple classification answers. We demonstrate its effectiveness for multi-view pedestrian detection. We estimate the marginal probabilities of presence of people at any location in a scene, given the responses of classifiers evaluated in each view. Our approach naturally takes into account both the occlusions and the very low metric accuracy of the classifiers due to their invariance to translation and scale. Results show our method produces one order of magnitude fewer false positives than a method that is representative of typical state-of-the-art approaches. Moreover, the framework we propose is generic and could be applied to any detection-by-classification task.
Conference: Computer Vision Theory and Applications - VISAPP , pp. 375-382, 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.
    • ...In [14], the POM framework is used in conjunction with people detectors...

    Rok Mandeljcet al. Fusion of non-visual modalities into the Probabilistic Occupancy Map f...

    • ...Approaches performing object detection/classification in multiple cameras at different viewpoints are also relevant to current work [3, 16]...

    Markus Enzweileret al. Integrated pedestrian classification and orientation estimation

    • ...Hence, there has been a considerable interest in object detection and tracking within multiple cameras networks [6, 10, 12, 14, 16, 17]...
    • ...For that purpose, they first apply change detection (e.g., [12, 16, 17]) or a fixed pre-trained classifier (e.g., [6]) to estimate the foreground likelihood of specific pixels...
    • ...Then, this information is fused exploiting the common ground plane by either estimating a score map (e.g., [6, 12, 16]) or by estimating axes intersections (e.g., [17])...
    • ...These problems could be alleviated by using centralized fusion methods (e.g., [6, 12], predicting the detections based on a probability map on the common top view of all cameras, or by running a motion-based verification step (e.g., [21, 23]...
    • ...The second data set (Forecourt Scenario) was thank worthy provided by the authors of [6]...

    Peter M. Rothet al. Multiple instance learning from multiple cameras

    • ...On the other hand, the works in [2], [7], [1] adopt a topdown approach...
    • ...The approaches proposed in this second category mainly differ based on the kind of generative model they consider (rectangle or learned dictionary), and on the way they decide about occupancy in each point of the grid (combination of multiple view-based classifiers in [2], probabilistic occupancy grid inferred from background subtraction masks in [7], and sparsity constrained binary occupancy map for [1])...

    Damien Delannayet al. Detection and recognition of sports(wo)men from multiple views

    • ...Other approaches use color information [16, 13] or classification [2] to obtain evidence about people presence in Figure 2: The ground plane grid used for the PETS sequence...

    Jerome Berclazet al. Evaluation of Probabilistic Occupancy Map People Detection for Surveil...

Sort by: