michals2 / tensorflex

TensorFlex

Image labeling tool for for creating object detection models

Frontend

View Architecture

• LocaleProvider
  • Provider
    • ConnectedRouter
      • App
        • Switch
          • AppHeaderContainer
            • AppHeader
          • Home
        • LabelCreator
          • LabelCreatorProgressContainer
            • LabelCreatorProgress
          • Switch
            • ModelExplorer
            • FileUploaderContainer
              • FileUploader
            • LabelSpecifier
            • LabelToolContainer
              • LabelTool
                • LabelToolNavBar
                • LabelSVG
            • Split Dataset
            • TrainModel

Reducers Architecture

• RootReducer
  • modelCreatorReducer
    • modelCreatorProgressReducer
    • fileUploaderReducer
    • labelToolReducer
  • myModelsReducer
  • routerReducer

State Architecture

• State
  • labelCreator
    • labelCreatorProgress
      • step: #
      • prevButtonDisabled: bool
      • nextButtonDisabled: bool
    • fileUploader
      • files: [ { fileName: "", url: "", dims: { height: #, width: # } } ]
    • labels
      • labels: [ { id: #, labelTypeId: #, dims: { varies depending on labelType } } ]
      • activeLabelId: #
      • nextAvailableLabelId: #
    • labelTypes
      • labelTypes: [ { id: #, name: "", shape: "", color: "", } ]
      • activeLabelTypeId: #
      • nextAvailableLabelTypeId: #
  • router
    • location
      • pathname: ""
      • search: ""
      • hash: ""
      • key: ""

Backend

• AWS Deep Learning

Things to improve

• Routing
  • Browser history and the model creator steps can get out of sync
• Frontend
  • upload files to browser
  • Check the uploaded files to ensure they are readable image types
  • Add ability to add labels to images
  • Add ability to create new labels
  • Add login and saving of images/labels/user credentials