This repo contains a starting point for a new TensorFlow application. It is built around Docker. When you first run it, it will download and compile TensorFlow for your CPU. If your CPU supports SSE4.2, AVX2 etc, it will run faster compared to the ready made TensorFlow bin.
It also contains a sample neural network to test your new TensorFlow container.
Requirements
- You need to have Docker installed
Run in root folder,
docker-compose build && docker-compose up -d
Login to the container,
docker exec -it ai /bin/bash -c "TERM=$TERM exec bash"
Go to /data folder and run
python tf.py
This is used as a starting point for machine learning projects. For this reason, Keras and some other libraries come pre-installed. You may remove them at your own discretion.
The model is trained with a handful of photos, that I personally shot to avoid copyright infringement stuff. You may improve the accuracy by adding more photos.