Why can't i use yolov8-seg weights downloaded from Roboflow for inference on my PC?

[NgocTrang]

I trained the instance segmentation model on Roboflow and downloaded the weight from that project. But I can not use it to call inference with YOLOv8 Segmentation model in Ultralytics, it gives the following error:

Meanwhile the weight I downloaded from the ultralytics homepage still works. Before that i downloaded Yolov8 detection model weight from Roboflow to detect object successfully.

It seems like Roboflow exports the segmentation model with a slightly different structure or pipeline which is causing some parts of Ultralytics to not recognize it correctly? Or is the error due to some other reason? Please help me!

• Project Type: Object dSegmentation
• Operating System & Browser: Linux
• Project Universe Link or Workspace/Project ID: danger_segmentation-vps6t

[Patrick_Nihranz]

Hi NgocTrang!

I think the issue you are running into here is that there is a version mismatch between the version of Ultralytics used to train the model on Roboflow and the version you have locally. I would suggest to check the version of Ultralytics you are running and then check the model info to determine if there is a mismatch. should just be print(model.info())