Pedestrian Detection Using R-CNN Object Detector

Abstract

Pedestrian detection continues to hold a significant role in the concept, analysis and function of computer vision. Deep learning techniques in pedestrian detection have demonstrated powerful results in recent experiments and research. In this paper a powerful deep learning technique of R-CNN is evaluated for Pedestrian detection on two different pedestrian detection datasets. The experiment involves the use of a deep learning feature extraction model along with the R-CNN detector. The deep learning feature extraction used is the Alexnet. Transfer learning is performed on the feature extraction model to adjust the weights of the convolutional neural networks to favour classification on the selected datasets. The R-CNN detector is then trained on the deep learning feature extraction model for pedestrian detection. The results of the experiments as evidently demonstrated, indicate some important truths about the performance of R-CNN detector on varying datasets.