I have been beating my head against the wall for about a month trying to get tensorflow to do optical image recognition without a pi camera.
It looks like it may boil down to CV2 which I have tried in vain to install on a raspberry 4 Pi with raspbian. I end up getting beat to death with dependencies.
So this is a question for anybody who has managed to get tensorflow image recognition to work without a raspberry Pi camera how did you do it?
Sub question.. how did you get CV2 to work?
And I released this message in a bottle to the Internet lol
This is a very interesting area that I have been trying hard to get to work as well. I basically wanted to run tiny-YOLO on a raspberry pi zero. Looking forward to following this topic and hope more info will flow. I will also update when I make progress.
Cheers!
I haven't abandoned this project it's on a roadmap for some things I really want to do. I just put it on the back burner for a bit while I knock out another project and lick my wounds from failing at CV2 LOL..
I'll post up if I have any successes as well. Good to see someone else on this track.
So I gave in and decided to get a Jetson nano after looking at the YouTube videos on how it works and what it can do. All of the things that were driving me crazy are pretty much included resources such as image recognition and such.. I can hardly wait for it to come in and play with it 🙂
I'm presently building the CUDA samples on my Nano, following the instructions in Bill's article. Guess I'll let it build while I sleep.
"Hardware eventually fails. Software eventually works." - Michael Hartung
I'm presently building the CUDA samples on my Nano, following the instructions in Bill's article. Guess I'll let it build while I sleep.
Cool... please build me a super duper render engine to handle Blender cycles! 😉
I actually managed to run two of the demos yesterday.. I did the thumbs up thumbs down demo and the regression demo with circles on your nose and eyes yada yada.. going to hit the bounding box thing today I think.. it's going really well so far 😀
Well the house didn't burn down from overheating, and the cpu has returned to normal temperature so I guess I'm good to go. This is way out of my comfort zone, glad I can't break anything playing with this stuff
😝
"Hardware eventually fails. Software eventually works." - Michael Hartung
Yay for not blowing up or overheating 🙂
I'm drinking from the fire hose with this. I'm finding out it's better to control it via SSH then to work on the desktop because when it's crunching numbers it is basically unresponsive for long periods and I have a twitchy reboot finger LOL
Onwards and upwards!
Fwiw I'm going to follow this up in another post since it's more nano than anything else.. hopefully a new post soon with pictures and stuff..
I just ordered a jetson nano v3. Can't wait to see the coming tutorials. Keep us posted here as to what you might discover 😀
I'm about three quarters of the way through the initial tutorials I made it through the setup and halfway through hello AI. Once I finish that I'm hoping to post some things up.
I strongly strongly suggest walking through all the tutorials and pushing the buttons along with them it's been extremely helpful for me and I clicked PC buttons for 25 years as a job