added video and poster

Author: Ceveloper

_data/authors_internal.yaml

@@ -18,7 +18,7 @@ abhisheksaroha:
 zhenzhangye:
   first_name: Zhenzhang
   last_name: Ye
-  website: https://cvg.cit.tum.de/members/yez
+  website: https://zhenzhang-ye.github.io/
 tarunyenamadra:
   first_name: Tarun
   last_name: Yenamandra
@@ -34,7 +34,7 @@ felixwimbauer:
 riccardomarin:
   first_name: Riccardo
   last_name: Marin
-  website: https://cvg.cit.tum.de/members/mricc
+  website: https://ricma.netlify.app/
 florianhofherr:
   first_name: Florian
   last_name: Hofherr

_publications/skeletondiffusion/content.html

@@ -1,28 +1,36 @@
-<!-- <section class="hero teaser">
-    <div class="container is-max-desktop">
-      <div class="hero-body">
-        <video width="100%" autoplay muted loop>
-          <source src="./assets/header_vid.mp4" type="video/mp4">
-        Your browser does not support the video tag.
-        </video>
-        <p class="content has-text-justified">
-          <i><b> Visual Odometry on KITTI. </b> Input image (top left), color-coded dense covariances (bottom left) and resulting trajectory (right). Color coding of the covariances is given by hue (size), orientation (color), and saturation (anisotropy). First sequence uses our supervised covariances, second sequence uses our unsupervised covariances.</i>
-        </p>
-      </div>
-    </div>
-  </section> -->
   <section class="hero is-small">
     <div class="hero-body">
       <div class="container">
+        <div class="video-container" style="margin: 1em 0;">
+          <iframe src="https://www.youtube.com/embed/W9GzdDXN41M" allowfullscreen style="width: 560px; height: 315px;"></iframe>
+        </div>
         <div class="columns is-centered has-text-centered">
-          <div class="column is-five-fifths">
-            <p class="content has-text-justified">
-                <img width="100%" src="./assets/teaser_image.png" />
-                <br><br>
-                <!-- a) We propose a symetric extension of the <a href="https://arxiv.org/abs/2204.02256" target="_blank">Probabiltistic Normal Epipolar Constraint</a>() (PNEC) to more accurately model the geometry of relative pose estimation with uncertain feature positions.
-                <br>
-                b) We propose a learning strategy to minimize the relative pose error by learning feature position uncertainty through differentiable nonlinear least squares (DNLS). This learning strategy can be combined with any feature extraction algorithm. We evaluate our learning framework with synthetic experiments and on real-world data in a visual odometry setting.  We show that our framework is able to generalize to different feature extraction algorithms such as SuperPoint and feature tracking approaches. -->
-            </p>
+          <div>
+            <style>
+              .video-container {
+                display: flex;
+                justify-content: center;
+                align-items: center;
+                width: 100%;
+                margin: 2em 0;
+              }
+              .video-container iframe {
+                width: 800px;
+                height: 450px;
+                max-width: 100%;
+                border: none;
+              }
+              @media (max-width: 900px) {
+                .video-container iframe {
+                  width: 100%;
+                  height: 56vw;
+                  min-height: 200px;
+                }
+              }
+            </style>
+            <!-- <div class="video-container">
+              <iframe src="https://www.youtube.com/embed/W9GzdDXN41M" allowfullscreen></iframe>
+            </div> -->
           </div>
         </div>
       </div>
@@ -56,4 +64,6 @@ <h2 class="title is-3">Coming Soon</h2>
         </div>
       </div>
     </div>
-  </section>
+  </section>
+
+  <!-- Probabilistic 3D human motion prediction aims to forecast multiple possible future movements from past observations. While current approaches often generate motions with undetected limb stretching and jitter. We introduce SkeletonDiffusion, a latent diffusion model that embeds an explicit inductive bias on the human body within its architecture and training. Our model is trained with a novel nonisotropic Gaussian diffusion formulation outperforming conventional isotropic alternatives. SkeletonDiffusion sets a new benchmark  across multiple evaluation metrics and datasets. -->

_publications/skeletondiffusion/index.md

@@ -40,6 +40,11 @@ links:
     - name: Code
       link: https://github.com/Ceveloper/SkeletonDiffusion/tree/main
       style: "bi bi-github"
+    - name: Video
+      link: https://www.youtube.com/watch?v=W9GzdDXN41M
+      style: "bi bi-youtube"
+    - name: Poster
+      link: https://cvpr.thecvf.com/virtual/2025/poster/33243
     # - name: Video
     #   link: # after video was created
     #   style: "bi bi-youtube"