magic-scissors 0.1.0

Generate synthetic data for computer vision projects using copy-paste context-augmentation.

Magic Scissors ✂️

Generate synthetic data for computer vision using copy-paste context augmentation.

Magic Scissors is available as a Python package and a web application.

Installation

To install Magic Scissors, run the following command:

pip install magicscissors

Quickstart 🚀

To use Magic Scissors, you need two datasets:

1. A dataset with object of interest;
2. A dataset with backgrounds on which objects of interest can be pasted.

Both datasets should be formatted as COCO JSON. You can convert data between formats using Roboflow.

from magic_scissors import MagicScissors

data = MagicScissors(
    dataset_size=100,
    min_objects_per_image=1,
    max_objects_per_image=3,
    min_size_variance=2,
    max_size_variance=5,
    annotate_occlusion=0,
    working_dir="./",
    upload_to_roboflow=False,
    roboflow_api_key="",
    roboflow_workspace="",
    roboflow_project="",
)

# load data from COCO JSON files
data.load_backgrounds_from_coco()
data.load_objects_of_interest_from_coco()

# load data from Roboflow
data.download_objects_of_interest_from_roboflow(
    dataset_url=""
)
data.download_backgrounds_from_roboflow(
    dataset_url=""
)

# generate dataset and save to directory
data.generate_dataset()

Contributing 🏆

We would love your help improving Magic Scissors! Please see our contributing guide to get started. Thank you 🙏 to all our contributors!

License

This project is licensed under an MIT license.