Roelof Pieters
Roelof Pieters
Looks like its common to export metadata in the gltf 'extras' field. So you could try something like (havent tried this but you get the idea): ``` let gltf; const...
it might be that your convolution kernel size is 4? whereas the original VGG uses 3, which should give more precise gfx.
You actually only need to enable Scene Semantics (available for iOS at extensions 1.40+). 1. update the arcore extensions to 1.40+ so you have the scene-semantics (semantic segmentation) also working...