Problems inferencing any Yolo model with the Webcam sample

Hi There,

I have trained a model using different Yolo versions using the Google colab notebook and I have downloaded my dataset from roboflow, trained it, and then uploaded my model weights to my Roboflow workplace.
When I run inference by using the “Try this model” and drop an image it works perfectly BUT if I try to use the real time webcam on browser it never opens and stays on “Loading Model Weights” screen forever; however if the model is trained by using the Roboflow Train option it works pretty fine so my web cam is fine and I have tried on different computers using different webcams.

Do Yolo models works using the webcam in realtime on my browser using your roboflow js script??
or why they might not be working?

I had raised the same issue just 3 days ago and it’s not yet solved for me as well. Hoping someone will come to our rescue. Also, I haven’t tried Roboflow Train option (only tried it after uploading my weights from colab)… thanks for letting me know it might work that way.

Yes, with their Train option it works but not with Yolo models for sone reason.
Also its hard to get their “train credits” as you cant buy them

@John_De_Leon @fahiar10 - Hi everyone, this bug has been previously reported, and filed. We are looking into it.

As a note, please search for similar reports prior to filing new threads. This is to ensure all users quickly find the information they need for bugs, and resolution.

I will report back when it is available. For now, if you still wish to use webcam or video inference, I have script here (“draw boxes” or “write text” under the video and stream columns): GitHub - roboflow/roboflow-computer-vision-utilities: Interface with the Roboflow API and Python package for running inference (receiving predictions) and customizing result images from your Roboflow Train computer vision models.

1 Like