Transfer image or video datasets from V7 Darwin to Roboflow (including annotations).

The V7Darwin-2-Roboflow toolkit is designed to automatically transfer datasets from V7 Labs' Darwin to Roboflow while retaining all annotations. Some common use cases include:

• Part of a dataset has been annotated in Darwin and needs to be merged with the rest of the dataset on Roboflow
• Switching from Darwin to Roboflow
• Missing Export Options in Darwin which Roboflow has (e.g. YOLOv8 for videos/video frames)

The code is flexible and can be utilized across various platforms, including Linux, MacOS, and Windows.

A Python environment running version 3.8 or later with all necessary dependencies is required. The dependencies can be installed by using the requirements.txt file and executing the following pip command in your terminal:

$ pip install -r requirements.txt  # Install all required packages

Additionally, a JSON and COCO export version (release) need to be created for your Darwin dataset in order for the toolkit to work. This can be done by following these steps:

1. Navigate to your dataset on the Darwin Website.
2. Click on Export Data.
3. Click the + Icon located at the top right of the Export Page.
4. Enter a name for the version. This name will be needed in the toolkit later.
5. Under Format, select Darwin 2.0 (JSON).
6. Under Items to Export, select All Complete.
7. Click on Export Items.
8. The release has now been created.

1. Navigate to your dataset on the Darwin Website.
2. Click on Export Data.
3. Click the + Icon located at the top right of the Export Page.
4. Enter a name for the version. This name will be needed in the toolkit later.
5. Under Format, select COCO.
6. Under Items to Export, select All Complete.
7. Click on Export Items.
8. The release has now been created.

You can either create a new Roboflow dataset or use an existing one in which the data should be uploaded. Make sure that you know the Project ID of this dataset.

Start by pasting your V7 Darwin and Roboflow API Keys and dataset information into v7darwin_2_roboflow.py instead of the placeholder values:

• V7 Darwin API Key
• V7 Darwin Dataset Name
• V7 Darwin JSON Release
• V7 Darwin COCO Release
• Roboflow PRIVATE API Key
• Roboflow Project ID

ATTENTION: Make sure that the V7 Darwin API Key has all necessary permissions enabled. Also, the Roboflow Project ID of a dataset does not always match the dataset name!

After that, you can start using the toolkit by executing the v7darwin_2_roboflow.py file. This will download all necessary data from Darwin, ensure that all annotations are retained, and upload the data into the specified Roboflow dataset.

def v7darwin_2_roboflow(
    video=False,
    remove_unannotated=False,
    overwrite_existing_data=False,
    clear_dataset_dir=False,
)

• video (bool, optional):

  • Indicates whether the V7 Darwin dataset consists of videos. Videos will be processed as individual frames. Defaults to False.

• remove_unannotated (bool, optional):

  • Specifies whether unannotated data should be removed before transferring the dataset to Roboflow. Defaults to False.

• overwrite_existing_data (bool, optional):

  • Determines whether existing data and annotations on Roboflow should be overwritten during the transfer process in case of duplicates. Defaults to False.

• clear_dataset_dir (bool, optional):

  • Indicates whether the dataset directory on the local computer should be cleared before downloading from V7 Darwin. Defaults to False.

# Image data:
v7darwin_2_roboflow()

# Image data (with removal of unannotated data):
v7darwin_2_roboflow(remove_unannotated=True)

# Video data:
v7darwin_2_roboflow(video=True)

# Video data (with removal of unannotated frames):
v7darwin_2_roboflow(video=True, remove_unannotated=True)

# Image data (with overwriting existing data/annotations):
v7darwin_2_roboflow(overwrite_existing_data=True)

If you find the toolkit useful for your work, please consider citing it.

MIT

For bug reports, feature requests, and contributions, feel free to use GitHub Issues. For other questions and discussions, you can contact me directly.