3. Usage Guide

Main concept of the Blueoil is quite simple: with just 4 steps, you can generate a neural network runtime which can run on an FPGA.

  1. Prepare a training dataset.

  2. Generate a configuration file.

  3. Train a network according to a config.

  4. Convert the trained network to an FPGA-ready format.

You could say Blueoil comes “with batteries included”, i.e. you don’t need to bother about things like defining your own network architecture, writing your own data augmentation, etc.