How to Install TensorFlow from Source on the Raspberry Pi

Update: Installing the latest version of TensorFlow on the Pi is now as simple as issuing these commands in the terminal:
"sudo apt-get install libatlas-base-dev"
"pip3 install tensorflow"

But this video still shows you how to build TensorFlow from source if you want to do it the hard way! It also gives some good experience on working with compiler tools.

Get a Raspberry Pi: https://amzn.to/2Iki3fb
Get a PiCamera: https://amzn.to/2rKxarh

If you have questions, I usually respond more quickly on Twitter, so send me a tweet @EdjeElectronics ! https://twitter.com/EdjeElectronics

I did this video using a Raspberry Pi 3 Model B running Raspbian Stretch. I also successfully built and installed TensorFlow on a Pi 3 Model B+. I believe it will also work on a Pi 2, but I haven't tried.

The sequel to this video instructs you on how to use TensorFlow's Object Detection API on the Raspberry Pi, which takes some extra setup: https://www.youtube.com/watch?v=npZ-8Nj1YwY

---- Command to initiate TensorFlow build ----
bazel build -c opt --copt="-mfpu=neon-vfpv4" --copt="-ftree-vectorize" --copt="-fomit-frame-pointer" --local_resources 1024,1.0,1.0 --verbose_failures tensorflow/tools/pip_package:build_pip_package

---- Link to steps in video ----
1:26 Step 1. Update Raspberry Pi & install packages
3:15 Step 2. Set up flash drive swap space
4:45 Step 3. Compile and install Bazel
8:40 Step 4. Compile and install TensorFlow
13:31 Step 5. Test out TensorFlow!

---- Link to SamJAbraham's GitHub guide ----
https://github.com/samjabrahams/tensorflow-on-raspberry-pi/blob/master/GUIDE.md
Guide re-used under Apache 2.0 license (http://www.apache.org/licenses/LICENSE-2.0)