OBB model deployment not working

[Guilherme_Augusto]

I’m trying to deploy my yolo26-obb trained on a custom dataset. I used the command :

version = project.version(2)
version.deploy("yolo26-obb", "/kaggle/working/runs/obb/train/weights/", "best.pt")

But I can’t use it to make inferences ( Oops something went wrong, there might be issues with this model) or use it to label assist (It doesn’t detect anything).

• Project Type: : Oriented Bounding Boxes
• Operating System & Browser: W11 / Chrome
• Project Universe Link or Workspace/Project ID: https://app.roboflow.com/solarpanels-td8ja/ir-solar-panels-rotated/
• Do you grant Roboflow Support permission to access your Workspace for troubleshooting? (Yes/No): Yes

[Bar_Shimshon]

Hi there team!

Checking this a bit further, it looks like the OBB (Oriented Bounding Box) deployment to Roboflow hosted inference is not supported yet — the model uploaded fine, but the server has no handler to run it. This is a missing feature.

Workaround options right now:

1. Self-host inference using the model weights directly with Ultralytics: YOLO("best.pt").predict(image)
2. Open a feature request at Issues · roboflow/inference · GitHub to get OBB support added

Let me know how it goes!

Bar Shimshon