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.
Prepare a training dataset.
Generate a configuration file.
Train a network according to a config.
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.