READ_ME

This repo provides basic code used to train a DL model on Google cloud ai platform to classify food brands. I used InceptionV3 Keras implementation and achieved an accuracy of ~90%.

Find :
	- data_preparation.py : The processing on images before training the model
		** Image rescaling (200*200 pixels)
		** Data augmentation (configurable rotation)
		** Split between training and validation
	- cmd : the google cloud SDK command to launch the job on GC.
	- pickle/ : the directory containing the prepared data (see data_preparation.py)
	- setup.py, trainer/ : Standard configuration for ai-platform (refer to google cloud documentation)

Not provided :
	- Training data (please contact me at oscar.barbier-darnal@capfi.fr if interested)