AI application development combat handwriting recognition application”
The application of handwritten recognition has been very popular, such as input method, character recognition in pictures, etc. But for most developers, how to achieve such an application will still feel helpless. This paper starts with a simple MNIST training model and starts handwriting recognition with everyone.
At the end of this tutorial, you will get a useful AI application, perhaps your first AI application. Although there is a large distance from the actual use (the specific gap will be analyzed behind the article), you will have a preliminary understanding of the AI application and have the ability to build a model that can be applied to the actual application.
- A computer using a win10 64 bit operating system
- Refer to the last blog AI application development campaign – from scratch to configure the environment. Training and exporting the MNIST model on the computer.
With the introduction of the previous article, you can get a model that recognizes a single handwritten number, and the accuracy of recognition will be 98%, or even more than 99%. So how do we use this model to build applications?
The general steps are as follows:
- To achieve a simple interface, the user will use the mouse or touch screen.
Input into a picture。
- The model that will be generated
PackingGet up and become a class with an open data interface.
- Take a picture of the input
Normalization，Become the format that the data interface can use.
- Finally, inference is used to figure out which figure the picture should be and display it.
Is it very simple?