Group Project for Distrubuted Systems and Cloud Computing.
Dependencies for Execution
- The models git repo for TensorFlow needs to be pulled into every worker and ps node in your system. It should be pulled into the same folder as the test_ps.py and test_worker.py files (located in server/googlenet and node_code repestively).
- Python 3.5.3.
- A virtual environment with TensorFlow, Paramiko, netifaces installed. See here for more details on installing and using virtualenv.
Instructions for Execution
- Run python default_discovery.py in the discovery folder if you are using your own LAN with Pis connected to deploy code to all the Pis on your network.
- Run python -m http.server in the client folder to start an HTTP server on the machine you intend to use an PS in tensorflow.
- Run python server.py in the server folder to start the Flask server that will send requests to TensorFlow
- Run python test_worker.py in /home/pi/cloud_computing/Parallel-Image-Processing/node_code/ on each of your Pis
- Load the webpage in client/index.html and upload your image, probabilities of what the image is should appear on the screen.