Clothing Classifier (Deep Learning - CNNs & Autoencoders)

• Tech Stack: Python (Keras, skLearn (T-SNE), Numpy, Seaborn, SciPy)
• Github URL: Project Link

This project applies deep learning techniques to the Fashion-MNIST dataset, utilizing both Convolutional Neural Networks (CNNs) for high-accuracy image classification and Autoencoders for unsupervised feature extraction and dimensionality reduction.

The CNN achieved a 92% validation accuracy, demonstrating strong performance in recognizing diverse clothing items, while the autoencoder successfully learned compact latent representations that enable visual exploration and provide features for downstream machine learning tasks.

This work highlights the versatility and effectiveness of deep learning models for both supervised and unsupervised analysis in real-world datasets, demonstrating how CNNs can achieve high-accuracy automated recognition of visual patterns and how autoencoders can uncover meaningful, compressed representations of complex data.

Through this project, we learned not only how to architect and tune modern neural networks for practical tasks, but also how these models can reveal insights that drive better decision-making in fields like retail, manufacturing, and beyond. Ultimately, these results underscore the transformative potential of deep learning to extract value from large, unstructured datasets and to power advanced analytics applications in industry.