Gaussian Opacity Fields

Improved Mip-Splatting with better geometry.

Web: https://niujinshuchong.github.io/gaussian-opacity-fields/
Paper: Gaussian Opacity Fields: Efficient and Compact Surface Reconstruction in Unbounded Scenes
Authors: Zehao Yu, Torsten Sattler, Andreas Geiger

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 27.55 0.874 0.115 1h 13m 32s 34.59 GB
bicycle 25.52 0.792 0.197 1h 11m 12s 37.49 GB
flowers 21.71 0.639 0.308 53m 26s 33.63 GB
treehill 22.44 0.643 0.320 57m 48s 33.73 GB
stump 26.99 0.798 0.223 52m 14s 33.82 GB
kitchen 31.23 0.925 0.158 1h 12m 30s 22.96 GB
bonsai 31.94 0.940 0.247 59m 18s 23.38 GB
counter 28.83 0.907 0.258 1h 5m 60s 12.64 GB
room 30.59 0.913 0.282 1h 9m 1s 23.70 GB
Average 27.42 0.826 0.234 1h 3m 54s 28.44 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.

Scene PSNR SSIM LPIPS (VGG) Time GPU Mem.
lego 35.56 0.982 0.021 19m 4s 2.88 GB
drums 26.17 0.955 0.045 17m 50s 2.84 GB
ficus 35.19 0.988 0.013 11m 10s 2.71 GB
hotdog 37.46 0.985 0.028 17m 52s 2.61 GB
materials 30.20 0.961 0.043 14m 48s 2.72 GB
mic 36.06 0.992 0.008 21m 56s 2.99 GB
ship 30.68 0.901 0.131 29m 60s 5.64 GB
chair 36.28 0.988 0.017 14m 46s 2.79 GB
Average 33.45 0.969 0.038 18m 26s 3.15 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 23.20 0.871 0.194 - -
ballroom 22.84 0.818 0.107 - -
courtroom 21.15 0.781 0.168 - -
museum 19.92 0.761 0.152 - -
palace 16.46 0.683 0.443 - -
temple 20.29 0.794 0.234 - -
family 22.31 0.875 0.084 41m 32s 35.16 GB
francis 24.76 0.901 0.158 35m 52s 12.30 GB
horse 23.73 0.881 0.092 38m 29s 24.41 GB
lighthouse 21.80 0.833 0.181 - -
m60 28.04 0.906 0.104 40m 9s 26.18 GB
panther 28.47 0.910 0.102 40m 13s 30.99 GB
playground 23.89 0.869 0.142 - -
train 19.69 0.796 0.164 - -
barn 25.72 0.866 0.140 - -
caterpillar 21.78 0.791 0.187 - -
church 19.65 0.775 0.208 48m 2s 36.45 GB
courthouse 19.60 0.726 0.346 35m 12s 12.82 GB
ignatius 20.34 0.769 0.162 43m 50s 34.97 GB
meetingroom 24.31 0.862 0.140 40m 29s 23.30 GB
truck 22.33 0.860 0.099 - -
Average 22.39 0.825 0.172 N/A N/A