Image Recognition Web App (Back-end)

This project builds a simple image recognition webapp where users can upload up to 10 images and get a recognition of what each image shows as a result.
In order to use the webapp, the user will have to login with a username and a password, after that he/she will be redirect to the upload page.

Check the front-end documentation for more info:
https://github.com/TD-team3/img-recognition-web-app-fe

Framework

The back-end project was built on Django. The server was hosted on imgrecognitionteam3.pythonanywhere.com.\

Project Structure

The Django project was structured in this way: a core server folder and two Django apps for implementing the main functionalities of the webapp: login and upload. The two apps are inside the authentication_app and upload_app directories.
Each app has its own view for handling Http requests and returning responses: they were called respectively login and upload.

OOP

The code was written following an OOP approach in order to better organize functionalities and make the project more testable and adaptable. Below are the main classes that were used:

• Authentication
• Token
• ImgRecognition
• ImgRecognitionHandler

Request-Response flow and Recognition Output

The used data-interchange format is JSON.

Authentication phase

The server receives username and password. In case of success, it returns a token which is valid for 3 hours and gives permission to access the upload page.

Upload phase

The server receives username and token as well a file of images (1 to 10).
In case of success (correct username, valid token, valid images' formats and sizes), it performs recognitions based on a pre-trained model (ResNet model trained on the ImageNet-1000 dataset) and after that it returns a JSON with a recognition string for each image name.