• DepthAI
  • Need some advice: Instance Segmentation model

Hi I would like to use an instance segmentation model like Mask R-CNN on regions with motion. I have tried Deeplabv3 but I find it to be quite un accurate... Does anyone know if such a model is available to use with oak-d pipeline? Or how should I go about converting such a model? I am very new and it all still feels a little bit daunting, though I am eager to learn 🙂

All tips/links/ and ideas would be greatly appreciated.

CorLangerak changed the title to Need some advice: Instance Segmentation model .

So CorLangerak I am not familiar with any instance segmentation models that could be run on-device. As instance-segmentation is usually a bit heavier. That said, a lot of these models are being optimized over time and then becoming tractable.

The PINTO0309 modelzoo is a good one to look through to see what is the latest:
https://github.com/PINTO0309/PINTO_model_zoo

Katsuya updates it so often that I am not up to speed on the latest.

And please feel free to bring this up in our Discord, as many of the users there are now past what I know in terms of what can be run on our platform:

https://discord.gg/EPsZHkg9Nx

For example geax in there implemented both of these:

He ported them from MediaPipe from Google to DepthAI. And so there might be some instance segmentation models in MediaPipe that could be ported as well.

Thoughts?

Thanks,
Brandon