Custom YOLOv11n-seg Weights Upload Successfully but Cannot Run Image Preview or Workflow

🤝 Community Help
1
Hi Roboflow Team,

I’m posting this because I’ve searched the forum and found several similar segmentation-related threads, but I haven’t seen a clear resolution to this specific issue yet.

Summary

  • Custom YOLOv11-n segmentation weights upload successfully and are recognized as instance segmentation models.
  • However, the uploaded segmentation models cannot run image preview or workflow execution.

This suggests the issue is specific to segmentation inference, not training or upload.

What Works

  • Uploading yolov11-n detection weights to Roboflow using the Python SDK
  • Running image preview and workflows with the uploaded detection weights successfully on Roboflow.
  • Uploading yolov11-n segmentation weights to Roboflow (Similar to detection)

What Fails

Any attempt to run inference (image preview or workflow) using custom YOLOv11-n segmentation weights.

image
image1177Ă—816 108 KB

image
image1813Ă—895 183 KB

Code

from roboflow import Roboflow

rf = Roboflow(api_key="key")
workspace = rf.workspace("kawadard2")

workspace.deploy_model(
    model_type="yolo11n",  # Type of the model #yolov11n: detect
    # yolov11n - yolov11n-seg - yolov11 - yolo11n
    model_path="./segment",  # Path to model directory
    project_ids=["danger_segment"],  # List of project IDs
    model_name="yolov11-segment-nano",  # Name for the model (must have at least 1 letter, and accept numbers and dashes)
    filename="seg.pt")  # Path to weights file (default)

Key Observation

  • Notably, these segmentation weights were downloaded from another segmentation project that was trained directly on Roboflow, rather than trained externally. This suggests the issue is unlikely to be caused by the training environment or custom export process.

  • Additionally, on the code side, I experimented with all commonly used model_type values for YOLOv11 segmentation, including:
    yolov11n, yolov11n-seg, yolo11n, yolo26n, and yolo26n-seg.

Questions

  • Is there a known limitation or compatibility issue with custom YOLOv11 segmentation models on Roboflow?

  • Are custom segmentation models expected to be fully supported for image preview and workflow execution, or is this an acknowledged gap?

  • If this is currently not supported, is there any recommended way to reuse a model that has already been trained in one workspace within another workspace, in order to avoid unnecessary re‑uploading the dataset and retraining of an identical model?

I’d really appreciate any clarification, confirmation, or guidance on this.
At the moment, it’s difficult to tell whether this is a known limitation or an unresolved issue, since detection works flawlessly while segmentation consistently fails.

Thank you very much for your help.

Best regards,
Trang

  • Project Type: Yolov11 segmentation model
  • Operating System & Browser:windows
  • Project Universe Link or Workspace/Project ID:workspace=“kawadard2”, project_ids=[“danger_segment”]
  • Do you grant Roboflow Support permission to access your Workspace for troubleshooting? (Yes/No):Yes