When using a Keypoint project, there are two options for training (Roboflow & Yolo), but neither seem to give good results when using “Try this model”.
Roboflow model detects tens of objects on the image (though I only have one object)
Yolo model detects none
Thoughts?
Currently, I’m exporting the Dataset and training on my machine with Yolo v8 and it’s working fine.
The issues you’re experiencing with the model predictions are due to the fact that our system does not yet support video inference for keypoint detection projects in the browser. We recommend testing your keypoint detection model with image inputs to achieve accurate results. We are currently working to disable video inputs for these projects and plan to support video inference for this model type in the future. Thank you for your understanding!
