Â
Â
Â
Â
Â
1. When a user sends us an image of a question, it goes to our domain - justace.space - before reaching the backend endpoint.Â
2. At the backend, the text in the image is first transcribed by MathPix, an Optical Character Recognition (OCR) service, before it is passed onto our neural network.
3. At this stage, the text is encoded for processing by a pre-trained MathBERT model, before it passes through our PyTorch model for categorization.
4. Once the model provides predicted topics as an output, those topics are used to lookup the relevant materials from our database, which are then returned to the user.
Â
The entire application is hosted on DigitalOcean, which is ideal for prototyping of smaller projects, such as JustAce.