This repo has the weights of resnet-18 model training on cifar-10 custom data, where some images are made upside down, and the goal is to predict the orientation of the image(0/1 classification task).