I’m making a monocular depth estimation app for autostereoscopic viewing like this. However, when I built this app for LumePad, the load was too high and it stopped after only one frame. You guys have a very nice technique for local depth estimation and consistent depth on Android with LriaPlayer and LeiaTube. I want to incorporate your real-time depth estimation technology into my app. can it? If not, I would like information about your depth estimation model. Is the model published on OSS? Is it your own research model?
Hey @acidys230! Thanks for coming by the forum! We’ve passed along your questions to the team. What type of app are you building? Cheers, Marlon - Community Manager
If this is for game content or real time 3D in an engine like Unity, you can use the color + depth shader I wrote and open sourced. Basically you just need access to the regular back buffer and the depth buffer and it can produce a stereo image pretty fast (around 1ms). The code is for anaglyph 3D in Godot, but would not be hard to port to other engines.
Hey Andres! Welcome! Thanks for chiming in here! Maybe this can help @acidys230? Would be fun to hack together sometime. LumePad2 Hackathon anyone???
Oops, I forgot to put a link.
I’m creating an app with Unity. I’m excited about your company’s depth estimation technology.
I would also be very happy if Unity provided a way to access the camera’s color map and depth map.
That looks really cool.