Train ResNet18 on Tiny ImageNet dataset (download from-http://cs231n.stanford.edu/tiny-imagenet-200.zip) with Adam as the optimizer for classification task. Plot curves for training loss, training accuracy, validation accuracy and report the final test accuracy. Here consider accuracy as top-5 accuracy. 1. Use CrossEntropy as the final classification loss function [10 marks] 2. Use Triplet Loss with hard mining as the final classification loss function [30 marks] 3. Use Central Loss as the final classification loss function [40 marks] Compare the performance of different models and analyze the results in the report.

Note - The code for ResNet18 architecture and the loss functions needs to be implemented from scratch. Directly importing from the library is not allowed and 0 marks will be awarded for that

Python code, that can be implemented in colab