Blog Post: Portable Computer Vision: Accelerating Tensorflow Lite with Edge TPUs on a Raspberry Pi
https://medium.com/@grepLeigh/portable-computer-vision-accelerating-tensorflow-lite-with-edge-tpus-on-a-raspberry-pi-84e318798ce9
Hi. Had a few challenges getting the make rpi-install step to complete, and eventually got to the point where I believe there is at least one broken link in the TASK [pip], which is reported as
fatal: [raspberrypi.local]: FAILED! => {"changed": false, "cmd": ["/home/pi/rpi-vision/.venv/bin/pip", "install", "-e", "."], "msg": "stdout: Looking in indexes: https://pypi.org/simple, https://www.piwheels.org/simple\nObtaining file:///home/pi/rpi-vision\nCollecting tensorflow@ https://github.com/leigh-johnson/tensorflow-community-wheels/blob/master/tensorflow-2.0.0b0-cp35-none-linux_armv7l.whl;platform_machine==\"armv7l\"\n\n:stderr: ERROR: HTTP error 404 while getting https://github.com/leigh-johnson/tensorflow-community-wheels/blob/master/tensorflow-2.0.0b0-cp35-none-linux_armv7l.whl;platform_machine==\"armv7l\"\n ERROR: Could not install requirement tensorflow@ https://github.com/leigh-johnson/tensorflow-community-wheels/blob/master/tensorflow-2.0.0b0-cp35-none-linux_armv7l.whl;platform_machine==\"armv7l\" from https://github.com/leigh-johnson/tensorflow-community-wheels/blob/master/tensorflow-2.0.0b0-cp35-none-linux_armv7l.whl;platform_machine==\"armv7l\" (from rpi-vision==0.1.0) because of error 404 Client Error: Not Found for url: https://github.com/leigh-johnson/tensorflow-community-wheels/blob/master/tensorflow-2.0.0b0-cp35-none-linux_armv7l.whl;platform_machine==%22armv7l%22\nERROR: Could not install requirement tensorflow@ https://github.com/leigh-johnson/tensorflow-community-wheels/blob/master/tensorflow-2.0.0b0-cp35-none-linux_armv7l.whl;platform_machine==\"armv7l\" from https://github.com/leigh-johnson/tensorflow-community-wheels/blob/master/tensorflow-2.0.0b0-cp35-none-linux_armv7l.whl;platform_machine==\"armv7l\" (from rpi-vision==0.1.0) because of HTTP error 404 Client Error: Not Found for url: https://github.com/leigh-johnson/tensorflow-community-wheels/blob/master/tensorflow-2.0.0b0-cp35-none-linux_armv7l.whl;platform_machine==%22armv7l%22 for URL https://github.com/leigh-johnson/tensorflow-community-wheels/blob/master/tensorflow-2.0.0b0-cp35-none-linux_armv7l.whl;platform_machine==\"armv7l\"\n"}
Thank you @knud! I'm going to update this to point to the official TensorFlow 2.0 wheels, now that 2.0 is GA.
Thanks @leigh-johnson . As you may well suspect, there was a way around the problem and I was able to get things working. Mostly my own issues being the first crack at TF for me. Looking forward to Part 2, though because I am helping students with making their own models, I really can't wait. Thanks!
I'm glad you were able to get this working! @knud
What age group are your students in? I'd love to hear more about the models they're trying to train as well.
So sorry about the delay on Part 2. I have a full-time job and this is just a hobby for me, but it brings me a lot of joy. ✨
I work mainly with university students, and sometimes with grades 3-8 via the university's outreach program (DiscoverE, U of Alberta). However, some of the uni students have their own mentoring clubs where they teach programming and robotics and other stuff. They have ambitions to push AI-ML stuff down to junior and high school.
Like you, it's kind of a hobby that will transition into more time as I semi-retire. It already takes more than 10% of my time. Like today, when I am madly checking Arduinos and RPis, and other HW as part of prepping my kits for my role as HW/FW mentor at HackEDbeta in Edmonton.
Keep up the Medium articles!