'Model Upload is Processing' for Hours When Deploying YOLOv8n-seg Model on Roboflow

Hi everyone,

I’m encountering an issue while trying to deploy a YOLOv8 segmentation model on Roboflow. Here’s what I’ve done so far:

Steps Taken:

  1. Trained the YOLOv8n-seg model on Google Colab.
  2. Created a new project on Roboflow and uploaded a fittitious image to create a dataset version, as I did not wish to upload my actual dataset.
  3. Used the following script on Google Colab to deploy the model:

python

Copia codice

from roboflow import Roboflow

rf = Roboflow(api_key="[API key]")
project = rf.workspace("delirio-annotazione-foto").project("ciccio74")
version = project.version("1")

version.deploy(
    model_type="yolov8n-seg",
    model_path="/content/gdrive/My Drive/roboflow_ciccio74/best.pt"
)

The deployment process has been stuck at the “Model upload is processing. This page will automatically update when it’s ready to use.” message for several hours.

Additional Information:

  • My internet connection is stable.
  • I used the specified version of ultralytics (8.0.196).
  • I tried creating a second version of the dataset and repeating the deployment process, but the issue persists.

Has anyone else encountered this issue? Any suggestions on how to resolve it?

Thanks for your help!

Domenico

Hey @DomenicoAndreucci

I believe the issue is coming from your model_type in the version.deploy function call. You should be putting yolov8-seg instead of yolov8n-seg.

  • yolov5, yolov5-seg

  • yolov7, yolov7-seg

  • yolov8, yolov8-seg, yolov8-cls, yolov8-pose

  • yolov9

  • yolov10

-See our docs for more info

1 Like

Hi @leo Thank you so much for your assistance with the model deployment issue I was facing. Your guidance on using the correct model type (yolov8-seg instead of yolov8n-seg) was invaluable, and I successfully deployed my model on Roboflow.

However, I am encountering a new challenge related to the use of this model in CVAT for auto-annotation. When I upload an image on Roboflow, the elements are predicted correctly by the model. But, when I use the same image for auto-annotation in CVAT, the model fails to predict any elements.

I have also additional questions:

  • Is it normal for models tested with images directly on Roboflow to accept images with dimensions exceeding 6000 pixels, while the same models require images to be resized to around 6000 pixels on the longer side for auto-annotation in CVAT?
  • Are there any specific settings or configurations that I need to adjust to ensure consistent behavior between Roboflow and CVAT?

Thank you very much for your continued support and help.

Hey @DomenicoAndreucci

For this question and others, although we do have an integration with CVAT, we don’t provide support for it since we don’t maintain it. I would recommend reaching out to any support options they have for more help regarding CVAT.

If you would like, you can use Label Assist to use your existing models to label within Roboflow.