https://github.com/parth-shastri/mnist_gradcam

The Grad CAM interpretability for MNIST dataset
https://github.com/parth-shastri/mnist_gradcam

Last synced: about 2 hours ago
JSON representation

The Grad CAM interpretability for MNIST dataset

• Host: GitHub
• URL: https://github.com/parth-shastri/mnist_gradcam
• Owner: parth-shastri
• License: mit
• Created: 2023-08-05T13:51:23.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2023-08-30T09:07:46.000Z (over 1 year ago)
• Last Synced: 2024-11-11T15:22:52.477Z (about 2 months ago)
• Language: Python
• Size: 28.3 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Project Title: MNIST Digit Recognition with Grad-CAM

## Description

This project demonstrates the implementation of Grad-CAM (Gradient-weighted Class Activation Mapping) for the MNIST dataset. Grad-CAM is a technique that helps visualize which parts of an image are contributing the most to a deep learning model's prediction.

In this project, we use a pre-trained model on the MNIST dataset to recognize handwritten digits. We then use Grad-CAM to highlight the regions in the input image that the model focuses on to make its prediction. The implementation includes both uploading images and drawing digits on a canvas for recognition and visualization.

Paper - [Grad-CAM: Visual Explanations from Deep Networks

via Gradient-based Localization](https://arxiv.org/pdf/1610.02391.pdf)

## Table of Contents

- [Installation](#installation)

- [Usage](#usage)

- [Demo](#demo)

## Installation

1\. Clone the repository:

   

    git clone https://github.com/your-username/mnist-grad-cam.git

    

    cd mnist-grad-cam

   

2\. Create the environment and install the required dependencies:

   

    

    python venv -n /path/to/environment

  

    source /path/to/environment/activate

  

    pip install -r requirements.txt

   

3\. Train your own model by making changes to model.py and run train.py:

    python train.py.

## Usage

1\. Directly run the Streamlit app to interact with the project:

    streamlit run main.py

2\. Open your web browser and navigate to the provided local URL (usually http://localhost:8501).

3\. Use the app to:

   - Upload an image of a handwritten digit for recognition and Grad-CAM visualization.

   

   - Draw any digit in freeform using the canvas on the UI.

     

   - See the recognized digit along with the Grad-CAM heatmap.

## Demo

A small dive into the UI and the usage of the app:

![](demo/demo.gif)

## License

This project is licensed under the [MIT License.](LICENSE)