Gaussian Splatting

Official Gaussian Splatting implementation extended to support distorted camera models. It is fast to train (1 hous) and render (200 FPS).

Web: https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/
Paper: 3D Gaussian Splatting for Real-Time Radiance Field Rendering
Authors: Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler, George Drettakis

Mip-NeRF 360

Mip-NeRF 360 is a collection of four indoor and five outdoor object-centric scenes. The camera trajectory is an orbit around the object with fixed elevation and radius. The test set takes each n-th frame of the trajectory as test views.

Scene PSNR SSIM LPIPS (VGG) Time GPU mem.
garden 27.34
Paper's PSNR: 27.41
0.866
Paper's SSIM: 0.868
0.124
Paper's LPIPS (VGG): 0.103
28m 58s 14.75 GB
bicycle 25.21
Paper's PSNR: 25.246
0.764
Paper's SSIM: 0.771
0.240
Paper's LPIPS (VGG): 0.205
28m 25s 14.90 GB
flowers 21.60
Paper's PSNR: 21.52
0.604
Paper's SSIM: 0.605
0.367
Paper's LPIPS (VGG): 0.336
20m 53s 10.01 GB
treehill 22.44
Paper's PSNR: 22.49
0.631
Paper's SSIM: 0.638
0.377
Paper's LPIPS (VGG): 0.317
20m 51s 11.36 GB
stump 26.58
Paper's PSNR: 26.55
0.771
Paper's SSIM: 0.775
0.250
Paper's LPIPS (VGG): 0.21
22m 30s 12.53 GB
kitchen 31.14
Paper's PSNR: 30.317
0.926
Paper's SSIM: 0.922
0.155
Paper's LPIPS (VGG): 0.129
25m 35s 9.75 GB
bonsai 32.20
Paper's PSNR: 31.98
0.941
Paper's SSIM: 0.938
0.254
Paper's LPIPS (VGG): 0.205
19m 15s 8.35 GB
counter 28.96
Paper's PSNR: 28.7
0.907
Paper's SSIM: 0.905
0.258
Paper's LPIPS (VGG): 0.204
21m 53s 8.55 GB
room 31.43
Paper's PSNR: 30.632
0.917
Paper's SSIM: 0.914
0.287
Paper's LPIPS (VGG): 0.22
22m 23s 10.05 GB
Average 27.43
Paper's PSNR: 27.20
0.814
Paper's SSIM: 0.815
0.257
Paper's LPIPS (VGG): 0.214
23m 25s 11.14 GB

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 35.70
Paper's PSNR: 35.78
0.982 0.019 5m 59s 3.07 GB
drums 26.15
Paper's PSNR: 26.15
0.953 0.044 6m 3s 3.07 GB
ficus 34.79
Paper's PSNR: 34.87
0.987 0.013 5m 22s 2.82 GB
hotdog 37.64
Paper's PSNR: 37.72
0.985 0.026 5m 49s 2.85 GB
materials 30.01
Paper's PSNR: 30.0
0.959 0.043 5m 28s 2.91 GB
mic 35.49
Paper's PSNR: 35.36
0.991 0.008 6m 11s 3.02 GB
ship 30.85
Paper's PSNR: 30.8
0.904 0.130 8m 13s 3.99 GB
chair 35.84
Paper's PSNR: 35.83
0.987 0.015 5m 45s 2.89 GB
Average 33.31
Paper's PSNR: 33.31
0.969 0.037 6m 6s 3.08 GB

Tanks and Temples

Tanks and Temples is a benchmark for image-based 3D reconstruction. The benchmark sequences were acquired outside the lab, in realistic conditions. Ground-truth data was captured using an industrial laser scanner. The benchmark includes both outdoor scenes and indoor environments. The dataset is split into three subsets: training, intermediate, and advanced.

Scene PSNR SSIM LPIPS Time GPU mem.
auditorium 24.13 0.871 0.193 10m 54s 4.98 GB
ballroom 24.07 0.824 0.101 19m 55s 9.08 GB
courtroom 23.12 0.790 0.165 17m 12s 9.02 GB
museum 20.92 0.764 0.160 20m 55s 11.38 GB
palace 19.63 0.736 0.350 11m 21s 5.85 GB
temple 20.85 0.806 0.222 10m 56s 5.13 GB
family 24.43 0.865 0.095 14m 60s 6.31 GB
francis 27.22 0.897 0.169 10m 20s 4.28 GB
horse 23.82 0.875 0.103 11m 35s 4.50 GB
lighthouse 22.11 0.843 0.156 12m 3s 6.09 GB
m60 27.84 0.901 0.113 13m 18s 6.80 GB
panther 28.32 0.908 0.107 13m 19s 7.01 GB
playground 25.37 0.848 0.170 15m 33s 7.44 GB
train 21.67 0.791 0.171 11m 56s 5.39 GB
barn 27.51 0.852 0.160 11m 9s 6.22 GB
caterpillar 23.38 0.791 0.190 11m 32s 5.89 GB
church 22.79 0.811 0.177 17m 15s 8.36 GB
courthouse 22.22 0.779 0.266 11m 8s 10.34 GB
ignatius 21.53 0.776 0.153 16m 36s 8.82 GB
meetingroom 25.19 0.866 0.141 12m 39s 5.50 GB
truck 24.25 0.853 0.108 15m 18s 7.47 GB
Average 23.83 0.831 0.165 13m 48s 6.95 GB

SeaThru-NeRF

SeaThru-NeRF dataset contains four underwater forward-facing scenes.

Scene PSNR SSIM LPIPS Time GPU mem.
Curasao 24.15 0.738 0.318 22m 26s 9.30 GB
Panama 23.68 0.749 0.250 22m 41s 8.43 GB
IUI3 18.86 0.677 0.390 18m 36s 4.26 GB
Japanese Gradens 18.54 0.752 0.242 16m 55s 3.96 GB
Average 21.31 0.729 0.300 20m 9s 6.49 GB