Semantic segmentation with webcam help

Has anyone gotten semantic segmentation running using a webcam? I’m working on project and any help or guidance would be greatly appreciated!

Yes, many time.
You can start from scratch

Or use a pretrained model like MobileNetV2

Then you can use the sample on TexturesToTensor to blit the webcam into a tensor, pass that into the model and then analyze the output

Thank you. I’ve gotten the ResNet DUC model running but the post processing seems tricky. Are there any resources you can point me to that I could analyze?

What is the output like?
I’m assuming it’s a classID per pixel?