High-Resolution Data-Driven Approach for Pedestrian Detection Using Deep Learning

Pedestrian detection is one of the primary objectives in the field of computer vision demonstrating versatile applications in various domains including pedestrian monitoring in surveillance systems and driverless automobile assistance. Since the surveillance cameras are set high to capture a wide range of areas, they produce small and noisy images. When these footage are given as input, the extracted features of the pedestrian images produce an inferior detection rate. To address the above issue, this research suggests a solution employing stationary wavelet dilated residual (SWDR) networks that significantly enhance edge details of the low-resolution images thus improving the efficiency of the pedestrian detection system.