Machine learning training data

 Oxford flower 102 example

Data is already split into train and validation folder. For an additional 3rd dataset, we will need to make the decision to split from train or valid, depends on the prevalence of data, per class.

Within the folders, there are 102 folders, corresponding to the 102 classes of flowers. Some folders have more data, some classes/folders have fewer. The folder name is the numeric value of the flower class. 

This is an example what the flower image data look like. 

Most modern deep learning libraries have the capability to load and organize image data from folders, named with the correct conventions.