https://github.com/xprithvi/lenet-5

LeNet-5, a convolution neural network (CNN), for digit recognition was replicated in PyTorch and trained on the MNIST dataset with feature visualisation via a gradient-based method.
https://github.com/xprithvi/lenet-5

artifical-intelligense cnn-classification convolutional-neural-networks jupyter-notebook machine-learning mnist neural-network pytorch

Last synced: about 1 month ago
JSON representation

LeNet-5, a convolution neural network (CNN), for digit recognition was replicated in PyTorch and trained on the MNIST dataset with feature visualisation via a gradient-based method.

• Host: GitHub
• URL: https://github.com/xprithvi/lenet-5
• Owner: xPrithvi
• License: mit
• Created: 2023-09-20T16:20:41.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2023-09-29T09:38:29.000Z (over 1 year ago)
• Last Synced: 2024-01-30T12:07:04.766Z (11 months ago)
• Topics: artifical-intelligense, cnn-classification, convolutional-neural-networks, jupyter-notebook, machine-learning, mnist, neural-network, pytorch
• Language: Jupyter Notebook
• Homepage:
• Size: 979 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE.txt

Awesome Lists containing this project

README

        
# LeNet-5 on Jupyter Notebook

[LeNet-5](/Lecun98.pdf), a convolution neural network (CNN), for digit recognition was replicated in PyTorch and trained on the [MNIST dataset]([LeNet-5/Lecun98.pdf](https://www.kaggle.com/datasets/hojjatk/mnist-dataset))

on Jupyter Notebook with feature visualisation via a gradient-based method. The rectified linear unit (ReLU) was used instead of the hyperbolic tangent function mentioned in the orginal paper. An accuracy of over 97% on the test dataset was 

achieved by model "LeNet-JFW4E".

## Future Improvements

- Hyperparameter tuning via Bayesian optimization.

- Implementation of the segmenter to allow for the extraction of multiple digits from noisy images.