Deep Learning Project: CNN, RNN and GAN

Overview

This project focuses on implementing basic deep learning models using PyTorch. It includes:

• CNN for image classification (CIFAR-10)
• RNN, LSTM and GRU for sequence learning
• GAN for image generation (Fashion-MNIST)

The main objective is to understand how these models work and observe their performance during training.

---

Project Structure

deep-learning-project/ │ ├── main.py ├── requirements.txt ├── README.md │ ├── outputs/ │ └── generated_images/

---

How to Run

1. Install the required libraries:

pip install -r requirements.txt

2. Run the program:

python main.py

---

Results

CNN:

• Trained on CIFAR-10
• Loss decreased over epochs

RNN, LSTM, GRU:

• LSTM and GRU performed better than RNN

GAN:

• Outputs improved over epochs

---

Outputs

outputs/generated_images/

---

Observations

• CNN works well for images
• LSTM/GRU better than RNN
• GAN training is unstable

---

Conclusion

This project helped in understanding CNN, RNN and GAN models practically.

---

Author

Susan Riona D'Souza