Lakonik / SSDNeRF

[ICCV 2023] Single-Stage Diffusion NeRF
https://lakonik.github.io/ssdnerf/
MIT License
432 stars 23 forks source link

Reconstruct a Mesh with texture using pictures #14

Open nihao-eng opened 1 year ago

nihao-eng commented 1 year ago

Hello, may I ask how long it will take to reconstruct the Mesh with texture using the picture?

Lakonik commented 1 year ago

Hi! Inference time can be found in section B.3. of the paper.

Under the reconstruction setting with the same batch size, a single guided DDIM step or Langevin step takes 0.21 sec, and a single outer finetuning step takes 0.28 sec (when K_in = 4). This sums up to around 23 sec for reconstructing a batch of 8 Cars (single-view), and 102 sec for reconstructing a batch of 8 Chairs (single-view) with additional Langevin steps. Once the triplane latent codes are sampled, neural rendering can be performed in real time to synthesize the output images.

To get textured meshes however, you need additional post-processing code to extract them from the nerfs, which is not supported in this codebase.

nihao-eng commented 1 year ago

Hi! Inference time can be found in section B.3. of the paper.

Under the reconstruction setting with the same batch size, a single guided DDIM step or Langevin step takes 0.21 sec, and a single outer finetuning step takes 0.28 sec (when K_in = 4). This sums up to around 23 sec for reconstructing a batch of 8 Cars (single-view), and 102 sec for reconstructing a batch of 8 Chairs (single-view) with additional Langevin steps. Once the triplane latent codes are sampled, neural rendering can be performed in real time to synthesize the output images.

To get textured meshes however, you need additional preprocessing code to extract them from the nerfs, which is not supported in this codebase.

Thanks for your reply, I will read your paper carefully and try to extract mesh with texture