search

Qengineering / RPi-image

Raspberry Pi 4 Buster 64-bit OS with deep learning examples
https://qengineering.eu/deep-learning-examples-on-raspberry-32-64-os.html
BSD 3-Clause "New" or "Revised" License
133 stars 21 forks source link

Need Bullseye v11 #15

Open MauiJerry opened 2 years ago

MauiJerry commented 2 years ago

The new v11 Bullseye with proper 64bit has lots of upgrades AND is very hard to get a working tensorflow-camera test running it would be REALLY nice if y'all made an image with that OS and examples (python especially)

Qengineering commented 2 years ago

I've started to build an image with Bullseye. It will take some time to install everything. End of this week probably.

MauiJerry commented 2 years ago

thanks, your site hs been very helpful getting it running on buster but bullseye changed a lot. the libcamera2 etc changes are latest stumble. mostly im looking at pose_estimation

Qengineering commented 2 years ago

Please temper your expectations! I write C++ to get the fastest possible apps. Python simply cannot compete with C++. Since picamera2 is almost a python project for now, I don't think I'll be using it in an example. Plus, it's a beta and the last thing I'm waiting for is an avalanche of issues as the software progresses.

MauiJerry commented 2 years ago

If i can get pose-estimate running on Bullseye64, i'll work in fortran. ;-)

and a c++ app would probably be faster, maybe even with tolerable latency.

MauiJerry commented 2 years ago

found the Buster 16gb pose app and it works! Gets about 5-7fps with a pi cam v1.5 (ir). I will still love a Bullseye 64 version of the image, but i'm gonna build off your C++ pose example and add sending Pose data via Open Sound Control (OSC) to other apps (TouchDesigner, maybe Unity)

Learner00100 commented 1 year ago

@Qengineering is rpi image based on Bullseye ready? Can you give the download link please? Buster have many problems while installing package required for me.

Qengineering commented 1 year ago

I am happy to announce the new Bullseye version.