[x] What resolutions are needed?
Since we only need a still from the camera, we will settle for the highest resolution to have the best chances at having image clarity, so UXGA (1600x1200)
[x] What apis / open source solutions exist?
OpenCV (This is what I will be using)
Edge Impulse (Better used when in need of ML)
[x] What are (three) different options of recognizing a remote you should consider?
The first natural option I thought of is Edge detection using OpenCV
I have seen online about Template matching and Feature matching, both feasible over OpenCV. I don't know the details, though I kept references to go back to if needed.
A (overkill) solution would be to use Edge Impulse by developing machine learning models, with support for image processing tasks using CNNs. This is not realistic as one of the counter arguments is that we don't have a very large set of button images
[x] Which is the one you want to go for in the first prototype?
Edge detection ! :))
Additional notes: In today's meeting we were able to see the camera's stream and we were able to capture (via /capture) a picture that was stored on the laptop.
[x] What resolutions are needed? Since we only need a still from the camera, we will settle for the highest resolution to have the best chances at having image clarity, so UXGA (1600x1200)
[x] What apis / open source solutions exist?
Additional notes: In today's meeting we were able to see the camera's stream and we were able to capture (via /capture) a picture that was stored on the laptop.