add IGS method

data_extraction/data_extraction.py

@@ -98,6 +98,8 @@ def parse_table_to_df(table, rotated=False):
     table = re.sub(r'\$\S+\$', '', table) #remove all inside $$
     table = re.sub(r'\\textbf{([^}]*)}', r'\1', table) #remove all \textbf{}
     table = re.sub(r'\\cellcolor{[^}]*}{([^}]*)}', r'\1', table) #remove all \cellcolor{}
+    table = re.sub(r'\\cellcolor\[[^}]*\]{[^}]*}', '', table) #remove celcolors with other backets
+    table = re.sub(r'\\multicolumn\{\d+\}\{[a-z]\}\{([\d.]+)\}', r'\1', table) # remove '\\multicolumn{1}{l}{12.5}'
 
     # Find the tabular environment
     rows = re.findall(r'\\begin\{tabular\}.*?\\end\{tabular\}', table, re.DOTALL)
@@ -166,7 +168,7 @@ def tex_to_pd(tables, sources_file):
                 if df is None:
                     continue
                 if source in pd_tables:
-                    pd_tables[source] = pd.concat([pd_tables[source], df], axis=1)
+                    pd_tables[source] = pd.merge(pd_tables[source], df, on="Method", how="outer")
                 else:
                     pd_tables[source] = df
             else:
@@ -270,7 +272,7 @@ def df_to_results_csv(pd_tables, sources_file, csv_tables):
         pd_tables[source] = pd_tables[source].loc[:, ~duplicates]
         #find "ours" column
         for row in range(len(pd_tables[source])):
-            if "ours" in pd_tables[source]["Method"][row].lower():
+            if not pd.isna(pd_tables[source]["Method"][row]) and "ours" in pd_tables[source]["Method"][row].lower():
                 # filter out submethods, everything after "ours", "" if only "ours"
                 submethod = re.search(r'(?i)ours([^}]*)', pd_tables[source]["Method"][row]).group(1)
                 #iterate through all columns and transfer values
@@ -287,8 +289,14 @@ def df_to_results_csv(pd_tables, sources_file, csv_tables):
                             #convert to Bytes
                             value = int(float(value) * 1024 * 1024)
                             metric = "Size [Bytes]"
-                        elif metric in ["PSNR","SSIM","LPIPS"]:
-                            pass
+                        # elif metric in ["PSNR","SSIM","LPIPS"]:
+                        #     pass
+                        elif "psnr" in metric.lower():
+                            metric = "PSNR"
+                        elif "ssim" in metric.lower():
+                            metric = "SSIM"
+                        elif "lpips" in metric.lower():
+                            metric = "LPIPS"
                         else:
                             continue
 
@@ -308,7 +316,7 @@ def df_to_results_csv(pd_tables, sources_file, csv_tables):
                         if len(row_index) == 0:
                             #"Method","PSNR","SSIM","LPIPS","Size [Bytes]","Data Source","Comment"
                             print("Adding new row for ", source, " ", submethod)
-                            result_tables[dataset_name].loc[len(result_tables[dataset_name])] = {"Method": source, "Submethod": submethod, "PSNR": "", "SSIM": "", "LPIPS": "", "Size [Bytes]": pd.NA, "Data Source": "", "Comment": ""}
+                            result_tables[dataset_name].loc[len(result_tables[dataset_name])] = {"Method": source, "Submethod": submethod, "PSNR": "", "SSIM": "", "LPIPS": "", "Size [Bytes]": "", "Data Source": "", "Comment": ""}
                             row_index = result_tables[dataset_name].index[
                                 (result_tables[dataset_name]["Method"] == source) & 
                                 (result_tables[dataset_name]["Submethod"] == submethod)

data_extraction/data_source.yaml

@@ -45,3 +45,12 @@ ye2023mathematical:
   url: "https://github.com/nerfstudio-project/gsplat/tree/main/examples/benchmarks/compression/results"
   is_csv: True
 
+wu2024implicit:
+  url: "https://www.arxiv.org/src/2408.10041"
+  is_csv: False
+  table_names: 
+    - tab:exp1
+    - tab:exp2
+  table_rotated: True
+  filename: "sec/3_method.tex"
+

methods.bib

@@ -165,4 +165,14 @@ @misc{ye2023mathematical
     primaryClass={cs.MS},
     url={https://docs.gsplat.studio/main/},
     shortname={gsplat},
+}
+
+@article{wu2024implicit,
+  title={Implicit Gaussian Splatting with Efficient Multi-Level Tri-Plane Representation},
+  author={Wu, Minye and Tuytelaars, Tinne},
+  journal={arXiv preprint arXiv:2408.10041},
+  year={2024},
+  url={https://www.arxiv.org/abs/2408.10041},
+  abstract={Recent advancements in photo-realistic novel view synthesis have been significantly driven by Gaussian Splatting (3DGS). Nevertheless, the explicit nature of 3DGS data entails considerable storage requirements, highlighting a pressing need for more efficient data representations. To address this, we present Implicit Gaussian Splatting (IGS), an innovative hybrid model that integrates explicit point clouds with implicit feature embeddings through a multi-level tri-plane architecture. This architecture features 2D feature grids at various resolutions across different levels, facilitating continuous spatial domain representation and enhancing spatial correlations among Gaussian primitives. Building upon this foundation, we introduce a level-based progressive training scheme, which incorporates explicit spatial regularization. This method capitalizes on spatial correlations to enhance both the rendering quality and the compactness of the IGS representation. Furthermore, we propose a novel compression pipeline tailored for both point clouds and 2D feature grids, considering the entropy variations across different levels. Extensive experimental evaluations demonstrate that our algorithm can deliver high-quality rendering using only a few MBs, effectively balancing storage efficiency and rendering fidelity, and yielding results that are competitive with the state-of-the-art.},
+  shortname={IGS}
 }

methods/wu2024implicit.md

@@ -0,0 +1,3 @@
+### Implicit Gaussian Splatting with Efficient Multi-Level Tri-Plane Representation
+
+This method introduces a hybrid representation for splatting-based radiance fields, where Gaussian primitives are separated into explicit point cloud and implicit attribute features. The attribute features are encoded using a multi-resolution multi-level tri-plane architecture integrated with a residual-based rendering pipeline. It employs a level-based progressive training scheme for joint optimization of point clouds and tri-planes, starting with coarse attributes and refining them with higher-level details. Spatial regularization and a bootstrapping scheme are applied to enhance the consistency and stability of the Gaussian attributes during training.

project-page/index_template.html

@@ -377,7 +377,7 @@ <h2 class="title is-3" style="margin-left: 10px; margin-right: 10px;">
                 <span class="author-block">
                     {{ paper.authors }}</span>
             </div>
-            {% if paper.published_at != "arXiv" %}
+            {% if not "arXiv" in paper.published_at %}
             <span class="rounded-bullet"><b>{{ paper.published_at }}</b></span>
             {% endif %}
             <br>

results/DeepBlending.csv

@@ -20,3 +20,5 @@ wang2024end,,29.35,0.895,0.277,4135678,,https://raw.githubusercontent.com/USTC-I
 wang2024end,,29.48,0.899,0.265,6706771,,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
 wang2024end,,29.59,0.901,0.256,10959799,,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
 wang2024end,Baseline,29.63,0.902,0.252,17998223,,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
+wu2024implicit,-Low,30.63,0.904,0.293,6647971,,https://www.arxiv.org/src/2408.10041,
+wu2024implicit,-High,32.33,0.924,0.253,8115978,,https://www.arxiv.org/src/2408.10041,

results/MipNeRF360.csv

@@ -20,6 +20,8 @@ wang2024end,,26.03,0.764,0.299,6162423,,https://raw.githubusercontent.com/USTC-I
 wang2024end,,26.50,0.784,0.268,9761150,,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
 wang2024end,,26.87,0.796,0.248,15351860,,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
 wang2024end,Baseline,27.05,0.802,0.239,23460290,,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
+wu2024implicit,-Low,27.33,0.809,0.257,13107200,,https://www.arxiv.org/src/2408.10041,
+wu2024implicit,-High,27.62,0.819,0.247,26633830,,https://www.arxiv.org/src/2408.10041,
 ye2023mathematical,,26.64,0.788,0.270,6916294,360000.0,https://raw.githubusercontent.com/nerfstudio-project/gsplat/main/examples/benchmarks/compression/results,
 ye2023mathematical,,26.88,0.796,0.256,8796870,490000.0,https://raw.githubusercontent.com/nerfstudio-project/gsplat/main/examples/benchmarks/compression/results,
 ye2023mathematical,-1.00M,27.29,0.811,0.229,16038022,1000000.0,https://raw.githubusercontent.com/nerfstudio-project/gsplat/main/examples/benchmarks/compression/results,

results/SyntheticNeRF.csv

@@ -15,3 +15,5 @@ wang2024end,,32.01,0.961,0.042,1016378,,https://raw.githubusercontent.com/USTC-I
 wang2024end,,32.56,0.964,0.038,1331576,,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
 wang2024end,,32.97,0.966,0.035,1838842,,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
 wang2024end,Baseline,33.12,0.967,0.034,2314380,,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
+wu2024implicit,-Low,33.36,0.971,0.036,1939865,,https://www.arxiv.org/src/2408.10041,
+wu2024implicit,-High,34.18,0.975,0.032,2852126,,https://www.arxiv.org/src/2408.10041,

results/TanksAndTemples.csv

@@ -21,6 +21,8 @@ wang2024end,,22.98,0.812,0.233,3737081,,https://raw.githubusercontent.com/USTC-I
 wang2024end,,23.14,0.823,0.214,5492385,,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
 wang2024end,,23.28,0.831,0.202,8017022,,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
 wang2024end,Baseline,23.34,0.835,0.195,12022710,,https://raw.githubusercontent.com/USTC-IMCL/RDO-Gaussian/main/results,
+wu2024implicit,-Low,23.70,0.836,0.227,8849981,,https://www.arxiv.org/src/2408.10041,
+wu2024implicit,-High,24.05,0.849,0.210,13107200,,https://www.arxiv.org/src/2408.10041,
 ye2023mathematical,,23.54,0.838,0.200,6875669,360000.0,https://raw.githubusercontent.com/nerfstudio-project/gsplat/main/examples/benchmarks/compression/results,
 ye2023mathematical,,23.62,0.845,0.188,8728572,490000.0,https://raw.githubusercontent.com/nerfstudio-project/gsplat/main/examples/benchmarks/compression/results,
 ye2023mathematical,-1.00M,24.03,0.857,0.163,16100628,1000000.0,https://raw.githubusercontent.com/nerfstudio-project/gsplat/main/examples/benchmarks/compression/results,