Fast Human Head and Shoulder Detection Using Convolutional Networks and RGBD Data

Wassim A. El Ahmar, Farzan Erlik Nowruzi, Robert Laganiere; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2020, pp. 106-107

Abstract


We introduce a new real-time approach for human head and shoulder detection from RGB-D data based on a combination of image processing and deep learning approaches. Candidate head-top locations (CHL) are generated from a fast and accurate image processing algorithm that operates on depth data. We propose enhancements to the CHL algorithm making it three times faster. Various deep learning models are then evaluated for the tasks of classification and detection on the candidate head-top locations to regress the head bounding boxes and detect shoulder keypoints. We propose three different models based on convolutional neural networks for this problem. Experimental results for different architectures of our model are discussed. We also compare the performance of our models to other state of the art methods in terms of accuracy of detections and computational cost and show that our proposed models are on par with the state of the art in terms of precision-recall of head detection and precision of shoulders detection, with the biggest advantage of our models being in terms of computation time. We also analyze the effect of adding the depth channel on the performance of the network.

Related Material


[pdf]
[bibtex]
@InProceedings{Ahmar_2020_CVPR_Workshops,
author = {El Ahmar, Wassim A. and Nowruzi, Farzan Erlik and Laganiere, Robert},
title = {Fast Human Head and Shoulder Detection Using Convolutional Networks and RGBD Data},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
month = {June},
year = {2020}
}