Install tensorflow on raspberry pi
There is a goode guide from Sam Abrahams at https://github.com/samjabrahams/tensorflow-on-raspberry-pi
I tried: Installing from pip (easy) and it worked great.
So instead of repeating Sams steps go to his github page.
After you installed tensor flow correctly you try it. (see https://github.com/tensorflow/tensorflow/tree/master/tensorflow/contrib/pi_examples)
- Get the tensorflow sources and dependencies
tensorflow/contrib/makefile/download_dependencies.sh
- Install libjpeg-dev if necessary
sudo apt-get install -y libjpeg-dev
- Download a pretrained tensorflow model
wget https://storage.googleapis.com/download.tensorflow.org/models/inception_dec_2015_stripped.zip
- Unzip the downloaded model into the path tensorflow/contrib/pi_examples/label_image/data/
- From the root of the TensorFlow source tree, run make -f tensorflow/contrib/pi_examples/label_image/Makefile to build a basic example.
- Run tensorflow/contrib/pi_examples/label_image/gen/bin/label_image --image=your_image.jpg