search

pmj110119 / RenderOcc

[ICRA 2024] RenderOcc: Vision-Centric 3D Occupancy Prediction with 2D Rendering Supervision. (Early version: UniOcc)
447 stars 25 forks source link

How to build the continuous SDF by predicted voxel-wise density and semantic? #6

Closed perisage closed 1 year ago

perisage commented 1 year ago

The volume feature map input to density mlp and semantic mlp is converted to voxel-wise density and semantic. But in the rendering process, use continuous density and semantic fields for any viewpoint render. So how do you build the continuous fields by the voxel-wise density and semantics? Thanks!

pmj110119 commented 1 year ago

Sorry for the late reply.

Through interpolation, specifically using the bilinear mode of F.grid_sample().

perisage commented 1 year ago

Got it! Thanks for your reply.