/DataLoop

Primary LanguagePython

DataLoop Assignment

Github Link

Part 1 (UI and preparation):

  1. Create a demo account on the Dataloop platform - https://console.dataloop.ai/
  2. Create a project
  3. Create a dataset
  4. Upload up to 10 images of your choice into the dataset (you can use Google Images or any free data source you may find)
  5. Create a basic Recipe for this dataset. Make sure to use Labels and Attributes that have relevance to the data presented.
  6. Annotate the images with tools of your choice. Try and use at least 2 types of annotation tools.

Part 2 (SDK):

  1. Install dataloop SDK
  2. Create a script that does the following:
  • a. Create a new dataset (if existing, get it)
  • b. Add three labels to the dataset recipe (class1, class2, and key)
  • c. upload directory with five images (Single upload to all items in the directory)
  • d. Add a UTM metadata to an item user metadata - collection time {“collected”:}
  • e. Add a classification of class1 to the first two of the images you uploaded.
  • f. Add a classification of class2 to the rest of the images you uploaded.
  • g. Add five random key points with the label “key” to one item.

About Script

The Script aims to explore the potential of the dataloop SDK. Following of the functionality can be used.

  • Login to Dataloop.ai
  • Get or create the project.
  • Get or create the dataset.
  • Upload the single images or the directory.
  • Add label and attributes.
  • ToDo
    • Add metadata to the items
    • Add classification

Usage

dlp = DataloopAI() #Initialize the class.
existing_projects = dlp.list_project() # List projects

References

Queries

  1. Can't understand how to add metadeta. https://sdk-docs.dataloop.ai/en/latest/
  2. Need more clarity on point e,f, g