Architecture¶
The api is based on 5 modules:
data: contains an internal representation of the dataset, classes for converting datasets, fiftyone dataset
_models: torch model, its hyperparams, functions for training
optimize: set of hyperparams, optuna optimizer
modification model: export and conversion of weights
logs: python logging
The Dash library was used for dashboard. It is based on components and callbacks on these component elements.