How to connect a neural network to a web application?

Question

Neural networks

Hello!
I have: a trained neural network for image classification in Py (h5) and a web application in TypeScript. The application has not yet been deployed, but if it is urgently necessary for the implementation of the task, then I will do it.
What I need: I need to connect a neural network to a web application.
As I suppose how it should work: the user enters the site, uploads an image, the neuron receives this image, classifies and sends the result of the classification.
Unfortunately, I don't know how to do this. Help me please!

Answer 1

TensorFlow Serving raises a web service with gRPC and REST api. With model h5.
The scenario is something like this. Write as for a regular site receiving a picture. Next, it must be brought to the desired size and counted per pixel values ​​and form JSON. Send this JSON with cURL to the TensorFlow Serving port address and get the response in JSON. Most likely it's a probability vector - depends on your model.
Understand which class the highest probability belongs to, return the response to the web client.

Answer 2

new to typescript and node in general, but I suspect that, like puff, it has the ability to run external shell scripts. That is, the algorithm is as follows: we load the file, run the Python script from the command line with parameters and the path to the file, and wait for a response from the python. We process the response in the node and give it to the user. Profit!