Machine Learning Assignments

This repository contains the implementation of two assignments focusing on various deep learning models, including CNNs, ResNet, RNNs, LSTMs, and Conditional GANs.

---

Assignment 1: CNN and ResNet Models

Tasks:

1. Fashion MNIST Classification:

   • Classify Fashion MNIST images into 3 categories: clothes, shoes, and others.
   • Use a custom CNN model.

2. Emotion Detection:

   • Train and fine-tune a ResNet-18 model for emotion detection:
     • From scratch: Train the ResNet-18 model with random initialization.
     • Fine-tuning: Use a pretrained ResNet-18 model for improved performance.

Objectives:

• Compare the performance of models trained from scratch vs. fine-tuned.
• Analyze classification accuracy and example predictions.

---

Assignment 2: SMS Spam Classification and Fashion MNIST cGAN

Tasks:

1. SMS Spam Classification & SMS completion:

   • Models:
     • RNN-based Classifier
     • LSTM-based Classifier
   • Objective: Compare the performance of RNN and LSTM models with a similar number of parameters. Provide example results.

2. Fashion MNIST Conditional GAN:

   • Model: Conditional GAN (cGAN):
     • Generator: Produces images based on a latent code and class label.
     • Discriminator: Classifies (Image, Label) pairs as real or fake.
   • Objective: Generate high-quality images for all 10 classes and showcase 4 examples per class.

---