Extract segmented image to features or geojson

So thanks to the amazing scripts and samples I was able to iterate over 1.5 million streetview type images with cityscapes and mapillary to produce panoptic segmented images. Absolutely amazing.

My question is… now what? I don’t even know what to search for. Somehow, from these georeferenced images, i need to extract features. such as x y coordinates (using depth estimate and angle, and source location from EXIF data) of the identified segments.

At a minimum, i would at least like to identify “picture contains xy and z classes” in some type of attribute file.

But i can’t even seem to get the search terms right? Can someone point me in the right direction?