Referenced from https://github.com/rockname/PhotoSelectionClassifier

PhotoSelectionClassifierSequenceDiagram.txt

actor User
participant App
collections Photos
database CoreData

User -> App: Open Photo Picker View
App -> Photos: Fetch photos from the Photos app
App <- Photos
User <- App: Show fetched photos
User -> App: Select photos and share to an album
App -> CoreData: Save selected photos
App <- App: Transit to Album View
App -> CoreData: Fetch selected photos
App <- CoreData
App -> Photos: Request image data
App <- Photos
User <- App: Show shared photos

group Training
    App -> Photos: Get photos in the shooting date period at the beginning and end of the selected photo list
    App <- Photos
    App -> App: Split the photos displayed on the album screen as selected and the others as unselected
    alt When either selected / not-selected does not meet the minimum number of 20
        App -> App: Do not start learning due to lack of training data
    else When there are 20 or more in both selected / not-selected
        group Preprocess
            App -> App: Align the number of selected / not-selected photos to the smaller number
            App -> App: Resize image to short side 300px
            App -> App: Save image data as training data in select / not-selected directory
        end
        group Train
            App -> App: Start learning by specifying the path of training data
            App -> App: When training is complete, compile and save the generated model
        end
    end
end

User -> App: Open Photo Picker View
User <- App: If there is a compiled model, show auto-select button
User -> App: Press the automatic selection button
App -> App: Get a compiled model
User <- App: Input the image data of each photo into the model and judge selected / not-selected