So, one thing I'd like to do here that I haven't seen is that I'd like to somehow attempt to validate the responses people are saying. First, to cut down on spam, and second, to give this some smarts.

The idea is:

  • Generate a QR code that contains some simple information for validation

  • Provide a space beside the QR code (or inside to help with boundary detection) to place the test device

  • Have the user take a picture of the device and the QR code

  • Use ML to determine the result through the picture

For one, it would be possible to use this statically - print the QR code once and people can just use it to submit pre-filled form information. Second, you can do this dynamically which I think I would rank higher in the sense that I can trust the result more.

Machine Learning

While at the very basic level, I'll just trust what people put in - it's Buzzword time! I'll be looking at using ML.NET, Tensor, and Blazor to assess the test result client-side without requiring people to interpret results. This should be possible through the only issue here might be a) differences in RAT tests, I only have experience in the one kind they sent home with the kids and b) I need source images to train the model.

If I generate a QR code that has space for the roughly 1x2 testing device, then it should be possible to train the model to look at the "result window" on the testing device.

Overall, it looks like all the RATs have similar physical features that could be useful. A small circle, where the droplets are placed, and a rectangle where the results are.

Still, some thinking to do here.