readme~

folder 'buildNET' contains several scripts, the important ones are:
                  'folder_stats.py' which allows you to summarize how many data you have
                  'builder/buildNET.py' which allows you to create the files train, val and test
                                        from your images folders.

folder 'data' contains the folders with the source images

folder 'models' contains the different models with which you can run caffe with our datasets

folder 'utils' contains docs related to the project, examples of other networks, and theory slides of IA

#################################################
#### DIFFERENT DATASETS
folder 'foodCAT' contains the files train, val and test that 'buildNET/builder/buildNET.py' creates

folder 'foodCAT_OLD' same as foodCAT, but with 2 classes duplicated in both datasets

#################################################
script 'caffeWrapper.py' caltulates different kinds of accuracy on differents TEST dataSets. (https://groups.google.com/forum/#!topic/caffe-users/NzKEWAFPPfI)
	To be as general as posible script, we need to fit some args before execute:
	modelName: Name of the model to use, in order to get the '.caffemodel' file (see definition of dict 'models' for more understanding)
	modelType: In order to get the model definition depending on which dataset we'll use (posible options so far, net_TEST or net_TEST_just_foodCAT, where both are dicts on this code)
	nameProbsLayer: The top name of the last innerProduct blob used in the model definition 
	nameAccuracyTop1: The top name of the Accuracy blob used in the model definition
	nameAccuracyTop5: The top name of the Accuracy blob used in the model definition with the parameter 'top_k: 5'

	Example of Usage: caffeWrapper.testAccuracy('foodCAT_googlenet_food101', 'net_TEST', 'loss3/classifier_foodCAT_googlenet_food101', 'loss3/top-1', 'loss3/top-5')

#################################################
WARNINGS (TODO: UPDATE THIS DESCRIPTION):
after use 'buildNET/builder/buildNET.py', delete the absolute path, just leave it like relative from the main folder TFG.