Fashion_MNIST_Classification

Classification of Fashion Labels using CNN model

1. Achieved 94% test accuracy for the CNN model by performing hyperparameter tuning to determine the optimal network architecture.
2. Identified mystery labels in the dataset by leveraging encodings from the intermediate layer of a CNN model by applying dimensionality reduction with PCA and employing K-means and DBSCAN clustering algorithms for unsupervised classification.
3. Analyzed feature extraction through dimensionality reduction using PCA and Autoencoder and determined that Autoencoder generated the most effective representation based on improved classification test accuracy with KNN.