Datasets:
Tasks:
Object Detection
Modalities:
Text
Formats:
parquet
Languages:
English
Size:
10K - 100K
License:
Magic The Gathering Card Detection Dataset
This dataset is dedicated to people wanting to build card detection models.
It will emulate MTG cards in random positions and provide the visible corners positions for each card as well as the direction to the next corner for each corner.
Example
It contains 10k 1024x1024 pictures in the train split and 3k in the test split.
Structure
Each row of the dataset contains:
- id: (int) image id
- image: (binary) The binary image
- annotation: (array) An array of corners representations:
- x: float [0, 1] x relative position in the image
- y: float [0, 1] y relative position in the image
- visible: bool Is the point visible or hidden by other cards
- angle: float [-PI, PI] angle of the vector going to the next corner
- corner_id: int [0, 1, 2, 3] which card corner (top left, top right, bottom right, bottom left)
- A string containg JSON data: all metadata associated with each card present in the frame if one wants to go further [rarity detection / frame types / artists / mana values / etc...]
Credits:
This dataset is based on other existing MIT Licensed dataset:
- MTG-Json
- Scryfall
This project is unofficial Fan Content permitted under the Fan Content Policy. Not approved/endorsed by Wizards. Portions of the materials used are property of Wizards of the Coast. ©Wizards of the Coast LLC.
- Downloads last month
- 33