Neural Net for Handwriting
Train and test networks with handwritten digits
Frank Zielen
iPhone Screenshots
Description
Train and test a 3-layer network with handwritten digits. Use embedded MNIST data or capture custom images via camera or finger input. View training progress and adjust network parameters. Export model weights and compare accuracy metrics.
This app allows you to train and test a 3-layer neural net with handwritten digits (0...9).
For this, you may use the embedded MNIST data base including 40.000 handwritten digits. Furthermore, you can teach the net with your own handwriting by using your smartphone camera or writing on the screen with your finger.
The app is based on and inspired by the example of Tariq Rashid given in his book "Make Your Own Neural Network". The app source code is on GitHub.