Question to TensorFlow JS community:
What use cases do you have for depth estimation in the browser via TensorFlowJS? Clearly a lot of AR/XR use cases but if you have specific examples in mind would love to hear more.
The image below shows some existing Google research models currently not available in JS to give you a feel of the type of data that could come back from such an API. Would this be useful to you too if it were available in JS?
If so, what features would you want to see from such an API to make results most useable to you?
Maybe you could benefit from hit test functionality in addition to the depth map that comes back. Or maybe having relighting capabilities out of the box for the scene? What would be on your wish list and why?
Do let me know your feedback