Source code for machine learning tutorials and examples used in Arm's ML developer space.
Raspbian Stretch already supports Tensorflow 1.9 as of December 2018. However installing could be a challange. So follow below tips to install tensorflow on Pi Zero W running Raspbian Stretch
sudo pip3 install --no-cache-dir tensorflow # --no-cache-dir will fix OutOfMemory problem sudo apt install libatlas-base-dev
If you ever get issue related to sh1 checksum, pay attention to what url being used for downloading and change command to "sudo pip install <.whl file>"
You can use python 2.7 (running python ) for commands like record (python record ..) but for running trained model use python3 Below 3 commands will save you from hassle of running python3 train.py
sudo apt-get install libhdf5
sudo apt-get install libhdf5-dev
sudo pip3 install h5py
Deploy a TensorFlow MNIST model with the Arm NN inference engine.
Explore gesture recognition with TensorFlow and transfer learning on the Raspberry Pi 4 Model B, Pi 3 and Pi Zero.
Train a convolutional neural network from scratch to recognize multiple gestures in a wide range of conditions with TensorFlow and a Raspberry Pi 4 Model B or Pi 3.
Deploy a Caffe CIFAR10 model on Arm Cortex-M CPUs.