https://sites.temple.edu/trail/files/2021/11/XieXinDamesIROS2021.pdf
architecture of our deep neural network. Our input data consists of three separate channels, one from lidar and two containing pedestr an data, each of which is an 80 80 array. These 3-channel input data are fed to the deep neural network consisting of 1 convolutional layer, 6 bottleneck r