The Fastest Pedestrian Detector in the West

We demonstrate a multiscale pedestrian detector operating in near real time (5 fps on 640x480 images) with state-of-the-art detection performance. The computational bottleneck of many modern detectors is the construction of an image pyramid, typically sampled at 8-16 scales per octave, and associated feature computations at each scale. We propose a technique to avoid constructing such a finely sampled image pyramid without sacrificing performance: our key insight is that for a broad family of features, including gradient histograms, the feature responses computed at a single scale can be used to approximate features responses at nearby scales. The approximation is accurate within an entire scale octave. This allows us to decouple the sampling of the image pyramid from the sampling of detection scales. Overall, our approximation yields a speedup of 10-100 times over competing methods with only a minor loss in detection accuracy of about 1-3% on the Caltech Pedestrian dataset across a w...

• Type: Journal
• Year: 2010
• Where: BMVC
• Authors: Piotr Dollár, Serge Belongie, Pietro Perona
• Keywords: BMVC 2010, Computer Vision, Image, Image Pyramid, Multiscale Pedestrian Detector