Deep Multi-camera People Detection

Deep architectures are currently the top performing methods for monocular pedestrian detection. Surprisingly, they have not been applied in the multi-camera set-up. This is probably in large part due to the lack of large-scale labeled multi-camera data-sets with overlapping fields of view. Our main contribution is a strategy in which we re-use a pre-trained object detection network, fine-tune it on a large-scale monocular pedestrian data-set, and train an architecture which combines multiple instances of it on a small multi-camera data-set. We estimate performance on both a new HD multi-view data-set, and the standard one - PETS 2009, on which we outperform state of the art methods.