diff --git a/Gemfile.lock b/Gemfile.lock index b4d788f..1d4e513 100644 --- a/Gemfile.lock +++ b/Gemfile.lock @@ -65,6 +65,7 @@ GEM webrick (1.8.1) PLATFORMS + ruby x86_64-linux DEPENDENCIES diff --git a/_site/assets/media/mhentropy_pdf.pdf b/_site/assets/media/mhentropy_pdf.pdf new file mode 100644 index 0000000..a4f2468 Binary files /dev/null and b/_site/assets/media/mhentropy_pdf.pdf differ diff --git a/_site/assets/media/mhentropy_poster.pdf b/_site/assets/media/mhentropy_poster.pdf new file mode 100644 index 0000000..cffeca0 Binary files /dev/null and b/_site/assets/media/mhentropy_poster.pdf differ diff --git a/_site/assets/media/mhentropy_slides.pdf b/_site/assets/media/mhentropy_slides.pdf new file mode 100644 index 0000000..8e01e4e Binary files /dev/null and b/_site/assets/media/mhentropy_slides.pdf differ diff --git a/_site/index.html b/_site/index.html index dedd71d..ebb9aa7 100644 --- a/_site/index.html +++ b/_site/index.html @@ -6,7 +6,7 @@ - MHEntropy: Multiple Hypotheses Meet Entropy for Pose and Shape Recovery + MHEntropy: Entropy Meets Multiple Hypotheses for Pose and Shape Recovery @@ -25,22 +25,23 @@

- MHEntropy: Multiple Hypotheses Meet Entropy for Pose and Shape Recovery + MHEntropy: Entropy Meets Multiple Hypotheses for Pose and Shape Recovery +

@@ -49,30 +50,37 @@

- National University of Singapore + + Computer Vision & Machine Learning Group, School of Computing, National University of Singapore +
  • - ICCV 2023 + IEEE/CVF International Conference on Computer Vision (ICCV) 2023

@@ -91,13 +99,32 @@

+

+ Abstract +

+

+ For monocular RGB-based 3D pose and shape estimation, multiple solutions are often feasible due to factors like occlusions and truncations. This work presents a multi-hypothesis probabilistic framework by optimizing the Kullback–Leibler divergence (KLD) between the data and model distribution. Our formulation reveals a connection between the pose entropy and diversity in the multiple hypotheses that has been neglected by previous works. For a comprehensive evaluation, besides the best hypothesis (BH) metric, we factor in visibility for evaluating diversity. Additionally, our framework is label-friendly – it can be learned from only partial 2D keypoints, such as visible keypoints. Experiments on both ambiguous and realworld benchmarks demonstrate that our method outperforms other state-of-the-art multi-hypothesis methods. +

+ +

+ Poster +

+ + +

+ Presentation +

+
+ +
+

Citation 

             
 @inproceedings{chenyang2023MHEntropy,
-  title={MHEntropy: Multiple Hypotheses Meet Entropy for Pose and Shape Recovery},
+  title={ {MHEntropy}: Entropy Meets Multiple Hypotheses for Pose and Shape Recovery},
   author={Chen, Rongyu and Yang, Linlin and Yao, Angela},
   booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
   year={2023}
diff --git a/_site/assets/media/WalkingAround_Joe_glamr_vs_hybrik.mp4 b/assets/_media/WalkingAround_Joe_glamr_vs_hybrik.mp4
similarity index 100%
rename from _site/assets/media/WalkingAround_Joe_glamr_vs_hybrik.mp4
rename to assets/_media/WalkingAround_Joe_glamr_vs_hybrik.mp4
diff --git a/_site/assets/media/basketball_glamr_vs_hybrik.mp4 b/assets/_media/basketball_glamr_vs_hybrik.mp4
similarity index 100%
rename from _site/assets/media/basketball_glamr_vs_hybrik.mp4
rename to assets/_media/basketball_glamr_vs_hybrik.mp4
diff --git a/_site/assets/media/favicon.ico b/assets/_media/favicon.ico
similarity index 100%
rename from _site/assets/media/favicon.ico
rename to assets/_media/favicon.ico
diff --git a/_site/assets/media/glamr_overview.png b/assets/_media/glamr_overview.png
similarity index 100%
rename from _site/assets/media/glamr_overview.png
rename to assets/_media/glamr_overview.png
diff --git a/_site/assets/media/glamr_res1.mp4 b/assets/_media/glamr_res1.mp4
similarity index 100%
rename from _site/assets/media/glamr_res1.mp4
rename to assets/_media/glamr_res1.mp4
diff --git a/_site/assets/media/glamr_res2.mp4 b/assets/_media/glamr_res2.mp4
similarity index 100%
rename from _site/assets/media/glamr_res2.mp4
rename to assets/_media/glamr_res2.mp4
diff --git a/_site/assets/media/glamr_res3.mp4 b/assets/_media/glamr_res3.mp4
similarity index 100%
rename from _site/assets/media/glamr_res3.mp4
rename to assets/_media/glamr_res3.mp4
diff --git a/_site/assets/media/glamr_sample.mp4 b/assets/_media/glamr_sample.mp4
similarity index 100%
rename from _site/assets/media/glamr_sample.mp4
rename to assets/_media/glamr_sample.mp4
diff --git a/_site/assets/media/glamr_teaser.mp4 b/assets/_media/glamr_teaser.mp4
similarity index 100%
rename from _site/assets/media/glamr_teaser.mp4
rename to assets/_media/glamr_teaser.mp4
diff --git a/_site/assets/media/running_glamr_vs_hybrik.mp4 b/assets/_media/running_glamr_vs_hybrik.mp4
similarity index 100%
rename from _site/assets/media/running_glamr_vs_hybrik.mp4
rename to assets/_media/running_glamr_vs_hybrik.mp4
diff --git a/assets/media/WalkingAround_Joe_glamr_vs_hybrik.mp4 b/assets/media/WalkingAround_Joe_glamr_vs_hybrik.mp4
deleted file mode 100755
index 4285730..0000000
Binary files a/assets/media/WalkingAround_Joe_glamr_vs_hybrik.mp4 and /dev/null differ
diff --git a/assets/media/basketball_glamr_vs_hybrik.mp4 b/assets/media/basketball_glamr_vs_hybrik.mp4
deleted file mode 100755
index cfd2ea6..0000000
Binary files a/assets/media/basketball_glamr_vs_hybrik.mp4 and /dev/null differ
diff --git a/assets/media/favicon.ico b/assets/media/favicon.ico
deleted file mode 100644
index 424df87..0000000
Binary files a/assets/media/favicon.ico and /dev/null differ
diff --git a/assets/media/glamr_overview.png b/assets/media/glamr_overview.png
deleted file mode 100755
index ba3aefb..0000000
Binary files a/assets/media/glamr_overview.png and /dev/null differ
diff --git a/assets/media/glamr_res1.mp4 b/assets/media/glamr_res1.mp4
deleted file mode 100755
index 4b6f526..0000000
Binary files a/assets/media/glamr_res1.mp4 and /dev/null differ
diff --git a/assets/media/glamr_res2.mp4 b/assets/media/glamr_res2.mp4
deleted file mode 100755
index e059ca1..0000000
Binary files a/assets/media/glamr_res2.mp4 and /dev/null differ
diff --git a/assets/media/glamr_res3.mp4 b/assets/media/glamr_res3.mp4
deleted file mode 100755
index 1af7c79..0000000
Binary files a/assets/media/glamr_res3.mp4 and /dev/null differ
diff --git a/assets/media/glamr_sample.mp4 b/assets/media/glamr_sample.mp4
deleted file mode 100755
index 716bd38..0000000
Binary files a/assets/media/glamr_sample.mp4 and /dev/null differ
diff --git a/assets/media/glamr_teaser.mp4 b/assets/media/glamr_teaser.mp4
deleted file mode 100755
index 1be87b1..0000000
Binary files a/assets/media/glamr_teaser.mp4 and /dev/null differ
diff --git a/assets/media/mhentropy_pdf.pdf b/assets/media/mhentropy_pdf.pdf
new file mode 100644
index 0000000..a4f2468
Binary files /dev/null and b/assets/media/mhentropy_pdf.pdf differ
diff --git a/assets/media/mhentropy_poster.pdf b/assets/media/mhentropy_poster.pdf
new file mode 100644
index 0000000..cffeca0
Binary files /dev/null and b/assets/media/mhentropy_poster.pdf differ
diff --git a/assets/media/mhentropy_slides.pdf b/assets/media/mhentropy_slides.pdf
new file mode 100644
index 0000000..8e01e4e
Binary files /dev/null and b/assets/media/mhentropy_slides.pdf differ
diff --git a/assets/media/running_glamr_vs_hybrik.mp4 b/assets/media/running_glamr_vs_hybrik.mp4
deleted file mode 100755
index ba4ec1a..0000000
Binary files a/assets/media/running_glamr_vs_hybrik.mp4 and /dev/null differ
diff --git a/index.html b/index.html
index f98308c..7525f97 100644
--- a/index.html
+++ b/index.html
@@ -14,7 +14,7 @@
     
     
     
-        MHEntropy: Multiple Hypotheses Meet Entropy for Pose and Shape Recovery
+        MHEntropy: Entropy Meets Multiple Hypotheses for Pose and Shape Recovery
     
     
     
@@ -33,22 +33,23 @@
     
     

         

-            MHEntropy: Multiple Hypotheses Meet Entropy for Pose and Shape Recovery
+            MHEntropy: Entropy Meets Multiple Hypotheses for Pose and Shape Recovery
+            
         

     

     

         

             
@@ -57,30 +58,37 @@ 

                     
-                    National University of Singapore
+                    
+                    Computer Vision & Machine Learning Group, School of Computing, National University of Singapore
+                    
                 
             
             
    
                 
  • 
-                    ICCV 2023
+                    IEEE/CVF International Conference on Computer Vision (ICCV) 2023
                 
    • 
             

             
         

     

@@ -99,13 +107,32 @@ 

             
         
 
+        

+            Abstract
+        

+        

+            For monocular RGB-based 3D pose and shape estimation, multiple solutions are often feasible due to factors like occlusions and truncations. This work presents a multi-hypothesis probabilistic framework by optimizing the Kullback–Leibler divergence (KLD) between the data and model distribution. Our formulation reveals a connection between the pose entropy and diversity in the multiple hypotheses that has been neglected by previous works. For a comprehensive evaluation, besides the best hypothesis (BH) metric, we factor in visibility for evaluating diversity. Additionally, our framework is label-friendly – it can be learned from only partial 2D keypoints, such as visible keypoints. Experiments on both ambiguous and realworld benchmarks demonstrate that our method outperforms other state-of-the-art multi-hypothesis methods.
+        

+
+        

+            Poster
+        

+        
+
+        

+            Presentation
+        

+        

+            
+        

+
         

             Citation
         

         
             
 @inproceedings{chenyang2023MHEntropy,
-  title={{MHEntropy}: Multiple Hypotheses Meet Entropy for Pose and Shape Recovery},
+  title={ {MHEntropy}: Entropy Meets Multiple Hypotheses for Pose and Shape Recovery},
   author={Chen, Rongyu and Yang, Linlin and Yao, Angela},
   booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
   year={2023}