DMTet Hybrid 3D Representation Quiz

Answers and explanations appear as you take the quiz.

1. What two ingredients does DMTet combine in its hybrid representation?
   • A deformable tetrahedral grid and a signed distance field
   • A point cloud and a voxel grid
   • Two stacked neural radiance fields
   • A pixel grid and an audio spectrogram
2. What does the signed distance value at each vertex indicate?
   • The vertex color
   • Whether the point is inside (negative) or outside (positive) the surface
   • The mesh file format
   • The camera position
3. How is the explicit triangle mesh extracted in DMTet?
   • By rasterizing a 2D image
   • By clustering a point cloud
   • By a differentiable Marching Tetrahedra layer where the SDF sign flips
   • By ray marching a NeRF only
4. What makes DMTet's surface extraction differentiable despite topology changes?
   • It freezes all weights during training
   • It avoids using any neural network
   • It only ever produces a single fixed shape
   • Gradients flow through the predicted SDF values and vertex position offsets
5. How does DMTet add fine geometric detail efficiently?
   • By coarse-to-fine subdivision of tetrahedra near the surface
   • By rendering at 8K resolution always
   • By increasing the audio sample rate
   • By adding more cameras
6. Which generative system is built on DMTet?
   • Whisper
   • NVIDIA GET3D
   • AlphaFold
   • Stable Diffusion 1.5 base
7. What two quantities does DMTet predict at each tetrahedral grid vertex?
   • A camera focal length and a color label
   • A triangle index and a texture filename
   • A signed-distance value and a position offset
   • A light intensity and a material name
8. Where does DMTet place an explicit surface vertex along a tetrahedron edge?
   • At the edge midpoint regardless of SDF values
   • At the grid vertex with the largest positive SDF
   • At a random point chosen after mesh extraction
   • At the linearly interpolated zero crossing between opposite SDF signs