This repo is a demo for multiclass classification for visual data and display various basic tasks such building a model and saving it, acquistion of data, training and evaluating the performance etc. Classifying handwritten digits using LeNet 5 architecture, written from scratch in PyTorch, is what I considered trivial task in machine learning.

The average acurracy is 98.63%.

Note: This is a small side project when I started fresh for machine learning around 3 years ago (2020).

I have included conda environment requirements and jupyter notebook for demo.