Machine learning multiple subjects

Hi I watched some time ago your video on smile and no smile. it was impressive
I am working on an outdoor robot and I want it at first to recognise tree trucks and open grass areas. So we can avoid a collision
So if I took photos of grassy areas and tree trunks in the 4 seasons and many different types I think it would work, Am I correct
Now if I then want it to learn other subjects like humans, dogs, pathways, fallen branches would it be a simple matter of taking more pictures of these individual subjects and adding them to different folders

Yes, however, don’t expect miracles from the camera in AI. You need to constrain the problem.

Thanks for your speedy response

Can you provide a little inside into “constraining the problem”

Of my list of things to recognise how would you rank the chances of a high detect rate for the following

I assume that you would do one group at a time

Obviously the more pics the better, but the minimum would be?



tree trunks

braches on ground



concrete pathways


last question
can you recommend a relativiely easy to read book on this subject