Welcome to my solution to the Hello World of Computer Vision.
I took several approches to the MNIST dataset problem and copied the files for those runs.
-
Digi Recognizer.ipynb
- This approch is a simple KNN classifier from Scikit Learn While
- Best score I was able to get with this was about 96.500%
-
Digit Recognizer - CNN.ipynb
- A very simple CNN scored 99.114%
-
Digit Recognizer - CNN LDS.ipynb
- This run involved a larger data set that I created by augmenting the MNIST data set
- While there are more effecent ways to create this data I wanted to play around with python a bit so did things a bit more manurally
- Scored 99.585%
-
Digit Recognizer - CNN LDS LR.ipynb
- This made some modifications to the network and changed from 30 epochs to 100
- Scored 99.628%
The neural networks were ran on a Vega Frontier Edition GPU using Metal by making use of PlaidML and Keras
Datasets used can be downloaded here