search

likojack / bnv_fusion

This repository implements our CVPR2022 paper "BNV-Fusion: Dense 3D Reconstruction using Bi-level Neural Volume Fusion"
MIT License
135 stars 18 forks source link

Question about interpolation in global-level fusion. #9

Closed GostInShell closed 1 year ago

GostInShell commented 1 year ago

Hi! Thanks for contributing the code of your excellent work!

I have a question regarding the interpolation strategy in the volume rendering of global-level fusion. I notice that here the alpha is derived by first estimating the alpha at the corner points and then interpolating. This is different from recent works that first interpolate features, and then query the alpha. https://github.com/likojack/bnv_fusion/blob/17a64d7ba8a74946ff826e0603c5ff7827196a70/src/models/sparse_volume.py#L792

Does this strategy show better performance than the feature intepolation one? Are there any specific considerations?

likojack commented 1 year ago

We used corner points because each feature voxel is independent. We use the local coordinate rather than the global coordinate of the entire scene, which is different from some recent works.