This dataset contains more than 1 million images covering 340 doodle categories. It contains grayscale images of doodles organized by category extracted from the Quick, Draw! dataset. Each image represents a hand-drawn sketch from a different category and is processed for machine learning tasks. The Quick, Draw! dataset has about 50 million images.

Dataset Contents

• image: Grayscale images of graffiti, each 255×255 pixels.
• category: 340 graffiti categories, each stored in its own directory.
• Total number of images: 1,020,000, each category contains 3,000 images.

Dataset structure

• Graffiti folder: Contains 340 subfolders, each representing a different graffiti category. Each subfolder contains 3,000 images.
• CSV file (master_doodle_dataframe.csv): Contains additional metadata about the image, including:
  • countrycode: The country code of the user who drew the graffiti.
  • drawing: The drawing data is in JSON format.
  • key_id: A unique identifier for each doodle.
  • recognized:: Boolean value indicating whether the scribble is recognized.
  • word: Category label (e.g. "traffic light").
  • image_path: The file path where the image is stored.