This assignment was done as part of the COL333 Course Requirements. The assignment was divided into 2 parts : Classification and Representational Learning. For instructions to run the code, refer to the run.ipynb files in the respective folders. The reports for each part contain analysis for classification and representational learning.
-
Part 1 : The model achieved an accuracy of
92.77%on the test dataset. The report contains details regarding the architecture of the neural network. Techniques like Batch Normalisation, Data Augmentation, Adam Optimiser, Cosine Annealing Learning Rate Scheduling, etc. were used to improve accuracy. -
Part 2 : The model achieved an accuracy of
75.75%and the structural similarity score of0.828. The report contains details regarding the implementation of the variation autoencoder (VAE).