Regresar

Cascade classifiers and saliency maps based people detection

Abstract:

In this paper, we propose algorithm and dataset for pedestrian detection focused on HCI and Augmented Reality applications. We combine cascade classifiers with saliency maps for improving the performance of the detectors. We train a HAAR-LBP and HOG cascade classifier and introduce CICTE_PeopleDetection dataset with images from surveillance cameras at different angles and altitudes. Our algorithm performance is compared with other approaches from the state of art. In the results, we can see that cascade classifiers with saliency maps improve the performance of pedestrian detection due to the rejection of false positives in the image.