Upload of cvat dataset

Help
#1

Hi,

I am trying to do custom object detection using Grove - Vision AI Module V2 and Edge Impulse.
I would like to use cvat, which is easy to manipulate, as it is very difficult to label a large number of images.
However, I don’t know how to upload the cvat dataset to Edge Impulse.

Has anyone tried this?
Could you please advise?

#2

Hi @matsujirushi

Although I have not got much experience in that I took a look at CVAT and built a test project to upload with CVAT format:

Export the Annotations from CVAT

  • Format Selection: Export the annotations from CVAT in a format we can support, such as COCO JSON or Pascal VOC XML. see more here: Uploader | Documentation
  • Download the Exported Files: Ensure that both the image files and their corresponding annotation files are downloaded and organized in a way that matches the expected directory structure.

2. Prepare the Dataset

  • Ensure your images and annotations are in the same directory and that each image has a corresponding annotation file (e.g., .xml for Pascal VOC or .json for COCO).

3. Uploading using Edge Impulse studio web ui:

  • Go to the “Data acquisition” page in Edge Impulse Studio.
  • Click the Uploader button to open the file uploader interface.
  • Choose the option to upload individual files or a folder containing your dataset.
  • Make sure to select both the image files and their corresponding annotation files when uploading.
  • Edge Impulse will try to automatically detect the format of the annotations (e.g., COCO JSON, Pascal VOC XML) and apply them accordingly.

See our uploader docs for more options on dataset upload via api / script:

Best

Eoin

1 Like