Transformers Deep Learning Project

Overview

This repository contains deep learning resources and code, primarily focused on Convolutional Neural Networks (CNNs) for image classification, as well as supporting reports and documentation for related coursework.

Visualizations

Transformer Architecture

Figure: Simple schematic of a Transformer model with 6 encoder and 6 decoder blocks, showing input, output, and connections.

Training and Validation Loss

Figure: Example of training and validation loss curves over epochs.

Training and Validation Accuracy

Figure: Example of training and validation accuracy curves over epochs.

Contents

• CNN_PA_Finalversion.ipynb: Jupyter notebook for a binary image classification task (dogs vs. cats) using a Convolutional Neural Network (CNN) with PyTorch. Includes data loading from Google Drive, model building, training, evaluation, and visualization of results.
• CNN_PA_Finalversion_deepniti.ipynb: Another version of the CNN notebook for the same dogs vs. cats classification task, with similar structure and requirements, possibly with different implementation details or experiments.
• DLreport.pdf: PDF report, likely summarizing the deep learning project, results, and findings. (Open with a PDF viewer for details.)
• CSE676_RNN.pdf: PDF document, likely related to coursework or reference material on Recurrent Neural Networks (RNNs). (Open with a PDF viewer for details.)

How to Use

1. Requirements:

   • Python
   • PyTorch
   • Jupyter Notebook
   • (Optional) Google Colab for running notebooks with Google Drive integration

2. Running the Notebooks:

   • Download the dataset from the provided Google Drive link (see inside the notebooks for URLs).
   • Mount Google Drive if using Colab, or place the data in the appropriate local directories.
   • Open either CNN_PA_Finalversion.ipynb or CNN_PA_Finalversion_deepniti.ipynb in Jupyter or Colab.
   • Follow the instructions in the notebook to train and evaluate the CNN model.

3. Reports:

   • Open DLreport.pdf for a summary of the project and results.
   • Open CSE676_RNN.pdf for additional reference on RNNs.

Notes

• The notebooks are designed for a binary classification task (cats vs. dogs) and require the dataset to be structured into train, val, and test folders.
• Ensure all dependencies are installed before running the notebooks.
• For any issues or questions, refer to the comments and instructions within each notebook.

---

This README was auto-generated based on the contents of the repository. For more details, please refer to the individual files.