This project implements a snow detection algorithm for pedestrian safety, using a custom convolutional neural network with spatial attention. It leverages both real-world and synthetic datasets, and combines predictions from VGG-19 and ResNet-50 models for enhanced accuracy.