Welcome to my Computer Vision project focusing on the MNIST dataset! This repository contains a Jupyter Colab notebook with implementations of three different models for digit classification.
-
In the notebook (
mnist_digit_classification.ipynb), I've implemented three different models for digit classification using the MNIST dataset.-
Custom CNN Model:
- Designed a Convolutional Neural Network (CNN) with two blocks, each consisting of Conv2d, MaxPool2d, and ReLU layers. The model is then followed by a classifier layer with Flatten and a linear layer to achieve the desired output shape.
-
LeNet-5 Architecture:
- Implemented the classic LeNet-5 architecture from scratch. This model has been historically significant in the development of Convolutional Neural Networks for image recognition tasks.
-
Modified LeNet-5 for Parameter Reduction:
- Tweaked the LeNet-5 architecture to significantly reduce the number of parameters while maintaining a high level of accuracy. This optimization aims to achieve a more computationally efficient model.
-
- Open the notebook (
mnist_digit_classification.ipynb) in Google Colab or Jupyter Notebook. - Run the cells sequentially to train and evaluate the models.
- Explore the impact of architecture changes on accuracy and model efficiency.
The project uses the MNIST dataset for digit recognition. You can find more information about the dataset here.
I plan to expand on this project by exploring more advanced architectures, experimenting with data augmentation techniques, and possibly incorporating transfer learning for improved performance on related datasets.
This project is licensed under the MIT License. Feel free to use and modify the code for your own learning and projects.
Special thanks to the MNIST dataset contributors and the PyTorch community for their valuable resources and frameworks.
Happy coding!