Merge pull request #14 from wyf0912/main

Add ContextGS table data, images and summary

data_extraction/build_html.py

@@ -661,7 +661,8 @@ def get_published_at(
                         pub = "SIGGRAPH"
                     elif "ACM International Conference on Multimedia" in pub:
                         pub = "MM"
-
+                    elif "Neural Information Processing Systems" in pub:
+                        pub = "NeurIPS"
                     pub = pub + " '" + entry["year"][-2:]
 
                 published_at[entry["ID"]] = pub

data_extraction/data_source.yaml

@@ -1,3 +1,7 @@
+wang2024contextgs:
+  url: "https://github.com/wyf0912/ContextGS/tree/main/results"
+  is_csv: True
+
 morgenstern2024compact:
   url: "https://github.com/fraunhoferhhi/Self-Organizing-Gaussians/tree/main/results"
   is_csv: True

methods/wang2024contextgs.md

@@ -0,0 +1,3 @@
+### ContextGS: Compact 3D Gaussian Splatting with Anchor Level Context Model
+
+This paper proposes the first autoregressive model at the anchor level for 3DGS compression. This work divides anchors into different levels and the anchors that are not coded yet can be predicted based on the already coded ones in all the coarser levels, leading to more accurate modeling and higher coding efficiency. To further improve the efficiency of entropy coding, a low-dimensional quantized feature is introduced as the hyperprior for each anchor, which can be effectively compressed. This work can be applied to both Scaffold-GS and vanilla 3DGS.

methods_compression.bib

@@ -1,3 +1,15 @@
+@misc{wang2024contextgs,
+  title={ContextGS: Compact 3D Gaussian Splatting with Anchor Level Context Model},
+  author={Wang, Yufei and Li, Zhihao and Guo, Lanqing and Yang, Wenhan and Kot, Alex C and Wen, Bihan},
+  booktitle={Neural Information Processing Systems},
+  year={2024},
+  shortname={ContextGS},
+  url={https://github.com/wyf0912/ContextGS},
+  abstract={
+      Recently, 3D Gaussian Splatting (3DGS) has become a promising framework for novel view synthesis, offering fast rendering speeds and high fidelity. However, the large number of Gaussians and their associated attributes require effective compression techniques. Existing methods primarily compress neural Gaussians individually and independently, i.e., coding all the neural Gaussians at the same time, with little design for their interactions and spatial dependence. Inspired by the effectiveness of the context model in image compression, we propose the first autoregressive model at the anchor level for 3DGS compression in this work. We divide anchors into different levels and the anchors that are not coded yet can be predicted based on the already coded ones in all the coarser levels, leading to more accurate modeling and higher coding efficiency. To further improve the efficiency of entropy coding, e.g., to code the coarsest level with no already coded anchors, we propose to introduce a low-dimensional quantized feature as the hyperprior for each anchor, which can be effectively compressed. Our work pioneers the context model in the anchor level for 3DGS representation, yielding an impressive size reduction of over 100 times compared to vanilla 3DGS and 15 times compared to the most recent state-of-the-art work Scaffold-GS, while achieving comparable or even higher rendering quality.
+  }
+}
+
 @misc{morgenstern2024compact,
       title={Compact 3D Scene Representation via Self-Organizing Gaussian Grids}, 
       author={Wieland Morgenstern and Florian Barthel and Anna Hilsmann and Peter Eisert},
@@ -11,6 +23,7 @@ @misc{morgenstern2024compact
       shortname={SOG},
 }
 
+
 @misc{lee2024compact,
       title={Compact 3D Gaussian Representation for Radiance Field}, 
       author={Joo Chan Lee and Daniel Rho and Xiangyu Sun and Jong Hwan Ko and Eunbyung Park},

results/DeepBlending.csv

@@ -24,6 +24,8 @@ papantonakis2024reducing,Baseline,29.63,0.902,0.249,18000000,1010000.0,https://r
 ren2024octreegs,Baseline,30.49,0.912,0.241,,112000.0,https://arxiv.org/pdf/2403.17898,
 taming20243dgs,Baseline,27.79,0.822,0.263,,270000.0,https://humansensinglab.github.io/taming-3dgs/docs/paper.pdf,
 taming20243dgs, (Big),30.14,0.907,0.235,,2810000.0,https://humansensinglab.github.io/taming-3dgs/docs/paper.pdf,
+wang2024contextgs,_lowrate,30.09,0.907,0.265,3654759,,https://raw.githubusercontent.com/wyf0912/ContextGS/main/results,
+wang2024contextgs,_highrate,30.41,0.909,0.259,6864817,,https://raw.githubusercontent.com/wyf0912/ContextGS/main/results,
 wang2024end,,29.35,0.895,0.277,4135678,318768.0,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
 wang2024end,,29.49,0.899,0.265,6706772,528719.0,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
 wang2024end,,29.59,0.902,0.257,10959800,887813.0,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,

results/MipNeRF360.csv

@@ -29,6 +29,8 @@ papantonakis2024reducing,Baseline,27.10,0.809,0.226,29000000,1460000.0,https://r
 ren2024octreegs,Baseline,28.05,0.819,0.217,,657000.0,https://arxiv.org/pdf/2403.17898,
 taming20243dgs,Baseline,27.29,0.799,0.253,,630000.0,https://humansensinglab.github.io/taming-3dgs/docs/paper.pdf,
 taming20243dgs, (Big),27.79,0.822,0.205,,3310000.0,https://humansensinglab.github.io/taming-3dgs/docs/paper.pdf,
+wang2024contextgs,_lowrate,27.62,0.808,0.237,13297458,,https://raw.githubusercontent.com/wyf0912/ContextGS/main/results,
+wang2024contextgs,_highrate,27.75,0.811,0.231,19308117,,https://raw.githubusercontent.com/wyf0912/ContextGS/main/results,
 wang2024end,,26.03,0.764,0.299,6162424,479851.0,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
 wang2024end,,26.50,0.784,0.268,9761150,776862.0,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
 wang2024end,,26.87,0.796,0.248,15351860,1237152.0,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,

results/SyntheticNeRF.csv

@@ -1,18 +1,18 @@
 Method,Submethod,PSNR,SSIM,LPIPS,Size [Bytes],#Gaussians,Data Source,Comment
-chen2024hac,-lowrate,33.24,0.967,0.037,1236009,99924,https://raw.githubusercontent.com/YihangChen-ee/HAC/main/results,
-chen2024hac,-highrate,33.71,0.968,0.034,1950692,143127,https://raw.githubusercontent.com/YihangChen-ee/HAC/main/results,
+chen2024hac,-lowrate,33.24,0.967,0.037,1236009,99924.0,https://raw.githubusercontent.com/YihangChen-ee/HAC/main/results,
+chen2024hac,-highrate,33.71,0.968,0.034,1950692,143127.0,https://raw.githubusercontent.com/YihangChen-ee/HAC/main/results,
 fan2024lightgaussian,Baseline,32.725,0.965,0.037,7838000,,https://arxiv.org/pdf/2311.17245,"Taken from Table 6, avg of per-scene results"
 lee2024compact,+PP,32.88,0.968,0.034,2799698,,https://arxiv.org/pdf/2403.14530,
 lee2024compact,Baseline,33.33,0.968,0.034,5809111,,https://arxiv.org/pdf/2403.14530,
-morgenstern2024compact, w/o SH,31.37,0.959,0.043,1978015,175871,https://raw.githubusercontent.com/fraunhoferhhi/Self-Organizing-Gaussians/main/results,
-morgenstern2024compact,Baseline,33.23,0.966,0.034,4132631,157322,https://raw.githubusercontent.com/fraunhoferhhi/Self-Organizing-Gaussians/main/results,
+morgenstern2024compact, w/o SH,31.37,0.959,0.043,1978015,175871.0,https://raw.githubusercontent.com/fraunhoferhhi/Self-Organizing-Gaussians/main/results,
+morgenstern2024compact,Baseline,33.23,0.966,0.034,4132631,157322.0,https://raw.githubusercontent.com/fraunhoferhhi/Self-Organizing-Gaussians/main/results,
 niedermayr2024compressed,Baseline,32.936,0.967,0.033,3686000,,https://arxiv.org/pdf/2401.02436,
 sun2024f3dgs,-CP-16,32.42,0.964,0.040,6354370,,https://arxiv.org/pdf/2405.17083,[N/T][N/P] PSNR 32.42 in Table 1 and 32.48 in Table 9
-wang2024end,,32.01,0.961,0.043,1016378,44811,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
-wang2024end,,32.56,0.964,0.038,1331577,64772,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
-wang2024end,,32.97,0.967,0.035,1838843,99988,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
-wang2024end,Baseline,33.12,0.967,0.035,2314380,132811,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
-wu2024implicit, low,33.36,0.971,0.036,1949555,157006,https://raw.githubusercontent.com/wuminye/ImplicitGS/main/results,
-wu2024implicit, high,34.18,0.975,0.032,2860954,157006,https://raw.githubusercontent.com/wuminye/ImplicitGS/main/results,
-xie2024mesongs, c3,32.96,0.968,0.033,3498725,207100,https://raw.githubusercontent.com/ShuzhaoXie/MesonGS/main/results,
-xie2024mesongs, c1,32.94,0.968,0.033,3873306,235531,https://raw.githubusercontent.com/ShuzhaoXie/MesonGS/main/results,
+wang2024end,,32.01,0.961,0.043,1016378,44811.0,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
+wang2024end,,32.56,0.964,0.038,1331577,64772.0,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
+wang2024end,,32.97,0.967,0.035,1838843,99988.0,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
+wang2024end,Baseline,33.12,0.967,0.035,2314380,132811.0,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
+wu2024implicit, low,33.36,0.971,0.036,1949555,157006.0,https://raw.githubusercontent.com/wuminye/ImplicitGS/main/results,
+wu2024implicit, high,34.18,0.975,0.032,2860954,157006.0,https://raw.githubusercontent.com/wuminye/ImplicitGS/main/results,
+xie2024mesongs, c3,32.96,0.968,0.033,3498725,207100.0,https://raw.githubusercontent.com/ShuzhaoXie/MesonGS/main/results,
+xie2024mesongs, c1,32.94,0.968,0.033,3873306,235531.0,https://raw.githubusercontent.com/ShuzhaoXie/MesonGS/main/results,

results/TanksAndTemples.csv

@@ -30,6 +30,8 @@ ren2024octreegs,Baseline,24.68,0.866,0.153,,443000.0,https://arxiv.org/pdf/2403.
 sun2024f3dgs,-CP-16,30.29,0.957,0.061,11471421,,https://arxiv.org/pdf/2405.17083,[N/T][N/P]
 taming20243dgs,Baseline,23.89,0.835,0.207,,290000.0,https://humansensinglab.github.io/taming-3dgs/docs/paper.pdf,
 taming20243dgs, (Big),24.04,0.851,0.170,,1840000.0,https://humansensinglab.github.io/taming-3dgs/docs/paper.pdf,
+wang2024contextgs,_lowrate,24.12,0.849,0.186,9902175,,https://raw.githubusercontent.com/wyf0912/ContextGS/main/results,
+wang2024contextgs,_highrate,24.29,0.855,0.176,12377181,,https://raw.githubusercontent.com/wyf0912/ContextGS/main/results,
 wang2024end,,22.98,0.812,0.234,3737081,263067.0,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
 wang2024end,,23.14,0.823,0.215,5492385,395042.0,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
 wang2024end,,23.28,0.831,0.202,8017022,597209.0,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,