Update index.html

index.html

@@ -211,7 +211,7 @@
  </td>
  <td align="center" width="200px">
  <center>
- <span style="font-size:16px"><a href="https://feng-hong.github.io/research/">Feng Hong</a><sup>2</sup></span>
+ <span style="font-size:16px"><a href="https://feng-hong.github.io/research/">Feng Hong</a><sup>1</sup></span>
  </center>
  </td>
  </tr>
@@ -254,7 +254,7 @@
  <tr>
  <td align="center" width="400px">
  <center>
- <span style="font-size:16px"><sup>2</sup>Shanghai AI Laboratory</span>
+ <span style="font-size:16px"><sup>2</sup>Shanghai Artificial Intelligence Laboratory</span>
  </center>
  </td>
  </tr>
@@ -314,7 +314,7 @@ <h2> Abstract </h2>
  <h2> Architecture </h2>
  </center>
  <p style="text-align:justify; text-justify:inter-ideograph;">
- <p><div style="text-align: center;"><img class="left" src="./resources/framework.png" width="600px"></div></p>
+ <p><div style="text-align: center;"><img class="left" src="./resources/framework.png" width="800px"></div></p>
  <p>
  <left>
  The framework of UniChest, which consists of two training stages. During the ``Conquer" stage, two modality encoders first project visual and textual representations into the common space with alignment, then feed them into the first transformer query networks for prediction. The multi-source common patterns are learnt as much as possible at this stage. During the ``Divide" stage, we freeze the modality encoders and squeeze the source-specific patterns via the MoE-QN module with the guidance of the enhanced supervised loss and the source contrastive learning.</left>
@@ -344,28 +344,21 @@ <h2>Results</h2>
  </center>
  <p>
  <left>
- We compare our best model with existing state-of-the-art approaches, to present a strong, yet simple baseline on video-text alignment for future research. 
- As shown in the table, on the challenging HT-Step task, 
- that aims to ground unordered procedural steps in videos, 
- our model achieves 46.7% R@1, leading to an absolute improvement of 9.3%, over the existing state-of-the-art (37.4%) achieved by VINA;
- On HTM-Align, which aligns narrations in the video, 
- our method exceeds sota model by 3.4%; 
- On CrossTask, where we need to align video frames and task-specific steps without finetuning, our method outperforms existing state-of-the-art approach by 4.7%,
- demonstrating our model learns stronger joint video-text representation.</left>
+ Performances under both in-domain and out-domain settings.</left>
  </p>
  <center>
- <p><img class="center" src="./resources/result.png" width="400px"></p>
+ <p><img class="center" src="./resources/result.png" width="800px"></p>
  </center>
- <br>
+<!--  <br>
  <hr>
  <center>
  <h2> Acknowledgements </h2>
  </center>
  <p>
  Based on a template by <a href="http:https://web.mit.edu/phillipi/">Phillip Isola</a> and <a
- href="http:https://richzhang.github.io/">Richard Zhang</a>.
- </p>
- <br>
+ href="http:https://richzhang.github.io/">Richard Zhang</a>. -->
+<!--  </p>
+ <br> -->
 </body>
 
 </html>