The project consists of two parts: basic task and advanced task. Basic task is a MNIST digit classifer implemented in neural network. And the advanced task is a handwriting digits recognizer.
Method: CNN
Accuracy: 99%
Model:
The advanced task is aiming to recognize hand-writing digits in photos. The whole process is divided into many stages, each of which is responsible for a sigle task.
Result:
Since all of those digits are written by my self, which is totally different from the style of MNIST. Plus some error in neural network. It only reaches 95% accuracy.