App for digit recognition developed using Tensorflow model, Django on the backend and React on the frontend.
To use the tool draw a digit from 0 to 9 on the drawable canvas.
when the mouse button is released, the image of the digit is sent in a body of a request to the backend.
Then the image is decoded, converted to greyscale, resized to 28x28 pixels etc. and finally passed to the model.
The previously built and loaded TF model makes a prediction of the digit and returns an array of probabilities.
This array is then passed back to the frontend as a response.
Based on this data and created useState hooks the text and graph are updated.