HUB: Upload CVAT annotated dataset and optimize automatically.

[agoransson]

Hello, I think the HUB is awesome! But I'm lacking a bit of an automation between my CVAT annotations to the ultralytics HUB datasets. It would really be nice to simply upload the exported CVAT annotated dataset to a zip and upload that to the HUB. Without having to organize it further.

• Accept CVAT zips
• Organize into optimized train/val/test ratios

Thanks for great tools!

[UltralyticsAssistant]

👋 Hello @agoransson, thank you for your kind words and for sharing your feedback about Ultralytics HUB 🚀! We're thrilled to hear that you're enjoying the tools and have ideas for improvement.

Your suggestion to streamline the process of uploading CVAT-annotated datasets directly to Ultralytics HUB is highly valuable. Enhancing automation, especially in dataset management, is something we continuously strive to improve. An Ultralytics engineer will review your feature request soon and provide further insights or updates.

In the meantime, if you have any examples or further details about how you envision this process working (e.g., specific formats or workflows), feel free to share them here. This will help us better understand and prioritize your request.

For additional resources, you can explore:

• Docs for general guidance.
• Ultralytics HUB Documentation for detailed info on HUB features and usage.

If you'd like to connect with the community or discuss this idea further, you can join us on:

• Discord 🎧 for real-time chat.
• Discourse for in-depth discussions.
• Subreddit to share and learn.

Stay tuned for updates, and thank you again for helping us make Ultralytics HUB even better! 😊

[glenn-jocher]

@agoransson thanks for the feedback! While direct CVAT zip uploads aren't currently supported, you can use CVAT's YOLO format export option and then follow our dataset preparation guide to create a HUB-compatible zip. We'll consider your automation suggestion for future improvements - please create a Feature Request in GitHub Issues to track this idea.

For current workflows, see our Datasets documentation explaining the required YAML structure and zip format.