I was hoping this technique would produce usable depth image, but I don't think it's quite there yet in realtime.
This Unity project I made following the steps from https://medium.com/xrpractices/monocular-depth-sensing-point-cloud-from-webcam-feed-using-unity-barracuda-d9f1496b5932 Feel free to check out the author @RKandas on github as well.
This project includes the TextureResizer, which from what I gathered was a dependency from tensorflow lite unity package.
Let me know if you have any issues using it. Here's the result of it https://www.loom.com/share/e31fd336aa0548d5bbc7c9170d787b3d