NeRF

Original NeRF method representing radiance field using a large MLP.

Web: https://www.matthewtancik.com/nerf
Paper: NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis
Authors: Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik, Jonathan T. Barron, Ravi Ramamoorthi, Ren Ng

Blender

Blender (nerf-synthetic) is a synthetic dataset used to benchmark NeRF methods. It consists of 8 scenes of an object placed on a white background. Cameras are placed on a semi-sphere around the object. Scenes are licensed under various CC licenses.

Scene PSNR SSIM LPIPS (VGG) Time GPU mem.
lego 32.86
Paper's PSNR: 32.54
0.964
Paper's SSIM: 0.961
0.049
Paper's LPIPS (VGG): 0.05
23h 32m 41s 10.25 GB
drums 25.12
Paper's PSNR: 25.01
0.926
Paper's SSIM: 0.925
0.087
Paper's LPIPS (VGG): 0.091
22h 56m 29s 10.25 GB
ficus 30.31
Paper's PSNR: 30.13
0.964
Paper's SSIM: 0.964
0.042
Paper's LPIPS (VGG): 0.044
23h 38m 35s 10.25 GB
hotdog 36.39
Paper's PSNR: 36.18
0.975
Paper's SSIM: 0.974
0.095
Paper's LPIPS (VGG): 0.121
23h 14m 14s 10.25 GB
materials 29.84
Paper's PSNR: 29.62
0.950
Paper's SSIM: 0.949
0.062
Paper's LPIPS (VGG): 0.063
23h 58m 42s 10.25 GB
mic 13.03
Paper's PSNR: 32.91
0.881
Paper's SSIM: 0.98
0.154
Paper's LPIPS (VGG): 0.028
23h 21m 46s 10.25 GB
ship 28.80
Paper's PSNR: 28.65
0.859
Paper's SSIM: 0.856
0.210
Paper's LPIPS (VGG): 0.206
23h 30m 51s 10.25 GB
chair 33.43
Paper's PSNR: 33.0
0.970
Paper's SSIM: 0.967
0.041
Paper's LPIPS (VGG): 0.046
23h 18m 37s 10.25 GB
Average 28.72
Paper's PSNR: 31.00
0.936
Paper's SSIM: 0.947
0.092
Paper's LPIPS (VGG): 0.081
23h 26m 30s 10.25 GB