https://github.com/sambhav/cvml-mnist

Various experimentations with the MNIST data set and ML models
https://github.com/sambhav/cvml-mnist

cnn computer-vision cv learning machine ml mnist neural-network recognition vision

Last synced: over 1 year ago
JSON representation

Various experimentations with the MNIST data set and ML models

• Host: GitHub
• URL: https://github.com/sambhav/cvml-mnist
• Owner: sambhav
• License: mit
• Created: 2017-09-02T14:12:40.000Z (almost 9 years ago)
• Default Branch: master
• Last Pushed: 2017-09-08T08:33:17.000Z (almost 9 years ago)
• Last Synced: 2025-01-24T05:41:18.257Z (over 1 year ago)
• Topics: cnn, computer-vision, cv, learning, machine, ml, mnist, neural-network, recognition, vision
• Language: Python
• Size: 11.9 MB
• Stars: 1
• Watchers: 6
• Forks: 2
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# cvml-mnist

Various experimentations with the MNIST data set and ML models

## Setup Instructions

* Setup virutal-env and setup dependencies

        

        virtualenv -p python3 venv

        source venv/bin/activate

        pip install -r requirements.txt

## Running the code

* To run the code without neural nets

        

        python src/mnsit_simple.py

* To run the code with neural nets

        

        python src/mnsit_cnn.py

## Saving results

* The code saves results in form of graphs as PNGs. The files are saved by the following scheme:

    * `{metric}_{model_type}_{batch_size}_{epochs}_{layers}{extra_params}.png`

    * `layers` and `extra_params` do not affect the experiment and are for file naming purposes only.

    * `batch_size` and `epochs` affect both the experiment and the filenaming scheme.

    * The only metrics saved are 'Accuracy' and 'Loss'

    * The graphs are plotted metric v/s epochs for both training and test data.

* Please change the parameters accordingly.