ultralytics / ultralytics

Ultralytics YOLO11 🚀
https://docs.ultralytics.com
GNU Affero General Public License v3.0
36.57k stars 7.05k forks source link

YOLO11 OBB + polygon annotations #19075

Open zxsk1974 opened 1 week ago

zxsk1974 commented 1 week ago

Search before asking

Question

Hi, currently I have YOLO11-OBB model trained on OBB dataset, but I would like to try to annotate some classes with polygon annotations instead of OBB. Hence few Qs:

  1. Does YOLO11-OBB supports polygon annotations?
  2. If not, should I convert all annotations to polygon and annotation format will be YOLO11, not YOLO11-OBB?
  3. Other recommendations?

Thanks!

Additional

No response

UltralyticsAssistant commented 1 week ago

👋 Hello @zxsk1974, thank you for your interest in Ultralytics 🚀! We recommend exploring the Docs where you can find detailed guidance on Ultralytics features and usage examples, including Python and CLI interfaces.

If this is a 🐛 Bug Report, kindly provide a minimum reproducible example to help us diagnose the issue effectively.

For your specific ❓ Question regarding YOLO11-OBB and polygon annotations:

We also encourage you to connect with the Ultralytics community for additional insights or ideas. Feel free to join us on Discord 🎧, Discourse, or our Subreddit.

Upgrade

Ensure you are running the latest ultralytics package with all requirements in a Python>=3.8 environment with PyTorch>=1.8. Upgrade using the following command:

pip install -U ultralytics

Environments

YOLO can be run in any of these verified environments (with all dependencies preinstalled, including CUDA/CUDNN, Python and PyTorch):

Status

Ultralytics CI

If this badge is green, all Ultralytics CI tests are currently passing. These tests ensure correct operation of all YOLO Modes and Tasks on macOS, Windows, and Ubuntu platforms.

This is an automated response 🤖. An Ultralytics engineer will review your question and provide further assistance soon. In the meantime, feel free to share any additional details that might help us help you! 🚀

Y-T-G commented 1 week ago

YOLO11-OBB uses the polygon formar but with 4 points. So you can use OBB formatted data to train segmentation model. The 4 points will define the mask. You can also maybe use polygon format with more than 4 points with OBB. I believe Ultralytics will then apply minAreaRect to get the oriented box from the polygon.

zxsk1974 commented 1 week ago

Hi @Y-T-G , so in case of mix annotated dataset wit OBB and Polgons, should I start with YOLO11-OBB or YOLO11-SEG model?

Y-T-G commented 1 week ago

segmentation if you don't want the angle

zxsk1974 commented 1 week ago

@Y-T-G , I need angle (oriented BB).

Y-T-G commented 1 week ago

Then you should use OBB