Self_Supervised_Learning

A ResNet18 model was trained using a self-supervised learning task in which the network was tasked with predicting the rotation of images. This approach enabled the network to learn the underlying structures and feature representations within the dataset. The pre-trained model was then applied to a downstream task—image classification on the CIFAR-10 dataset. Fine-tuning the network resulted in an accuracy of 80%.