Collecting images to train a target (and noise) in Edge Impulse seems quite arduous. Anybody have experience in using a photography turntable to facilitate the taking of multiple images? A logical stepper motor task.
Is there a particular thing you are concerned about? What you want to do sound like it will work.
I was watching the very clear Digi-Key Youtube video on finding a particular LEGO block (target) https://youtu.be/6wIswIpSw04. It seems that there must be a better way of taking the pictures to identify the target than just randomly. From my understanding, you want to take a lot of pictures (ideally 100’s or 1000’s)) - please add or comment on my assumptions below
- at different angles
- in different light conditions
- different degrees of occlusion (not sure about this one)
- insert more thoughts here
If what I am trying to track is like a LEGO brick (a small, motionless unit that can be isolated ), it seems the photography turntable can be an asset.
Am I correct that if I built my own turntable, I can both code the control the motor of the turntable and the taking of the picture all from the OpenMV Cam H7+?
Yep, there’s not really a lot of risk in this. We have motor driver code and it works.