This package implements a neural network trained on MNIST dataset to detect handwritten numbers.
The MNIST dataset of handwritten digits has a training set of 60,000 examples, and a test set of 10,000 examples:
mnist_train.csv is pretty big and Github cannot version control it. So in order to run training you have to unzip the mnist_train_csv.zip file in the same folder where it is currently placed.
To train the network and update the weights from the DetectNumbersCore folder run:
python Sources/TrainCore.py
And wait for some time depending on your hardware.