I am working on project where I has to execute Google Object detection API on a Raspberry Pi 3. The object detection api requires tensorflow version 1.4 or higher. Currently tensorflow is not developed for 32 bit operating systems. So how can I use tensorflow in raspberry pi. I try to use this project but since it has a older version of tensorflow it won't work.
- Is it possible to install 64 bit operating system on raspberry pi 3 and run normal tensorflow from official site? If 64 bit processor used is arm architecture would be a problem to run tensorflow?
- Is there any specific tensorflow wrapper or something that can be directly use in raspberry pi?
Please help me with available options for my problem. Thanks.
2 Answers 2
Binaries (32-bit) for TensorFlow 1.4 can be found from one of the tickets in the tensorflow-on-raspberry-pi project that you've linked: https://github.com/samjabrahams/tensorflow-on-raspberry-pi/issues/104
You'll likely have to install those *.whl files on Stretch, even though the main releases are usuallly built for Jessie. Thanks Chris for mentioning this in another question.
I would check out this repo here https://github.com/romilly/rpi-docker-tensorflow and it's probably what solves the issue temporary.