1

README.md

@@ -15,7 +15,7 @@
     </video> -->
     <img src="./assets/Teaser.png" alt="example" width=100%>
     <br>
-    <em>TLDR: We provide a novel dataset for extrapolated urban view synthesis, consisting of 90,810 frames across 345 videos. This dataset provides ground truth images for qualitative testing, addressing missing data issues. Additionally, we introduce a structured evaluation framework with categorized difficulty levels, facilitating robust assessment of extrapolated view synthesis.</em>
+    <em style="text-align: left;">TLDR: We provide a novel dataset for extrapolated urban view synthesis, consisting of 90,810 frames across 345 videos. This dataset provides ground truth images for qualitative testing, addressing missing data issues. Additionally, we introduce a structured evaluation framework with categorized difficulty levels, facilitating robust assessment of extrapolated view synthesis.</em>
 </p>
 
 
@@ -24,6 +24,8 @@
 
 Photorealistic simulators are essential for the training and evaluation of vision-centric autonomous vehicles (AVs). At their core is Novel View Synthesis (NVS), a crucial capability that generates diverse unseen viewpoints to accommodate the broad and continuous pose distribution of AVs. Recent advances in radiance fields, such as 3D Gaussian Splatting, achieve photorealistic rendering at real-time speeds and have been widely used in modeling large-scale driving scenes. However, their performance is commonly evaluated using an interpolated setup with highly correlated training and test views. In contrast, extrapolation, where test views largely deviate from training views, remains underexplored, limiting progress in generalizable simulation technology. To address this gap, we leverage publicly available AV datasets with multiple traversals, multiple vehicles, and multiple cameras to build the first <strong>E</strong>xtrapolated <strong>U</strong>rban <strong>V</strong>iew <strong>S</strong>ynthesis (EUVS) benchmark. Meanwhile, we conduct quantitative and qualitative evaluations of state-of-the-art Gaussian Splatting methods across different difficulty levels. Our results show that Gaussian Splatting is prone to overfitting to training views. Besides, incorporating diffusion priors and improving geometry cannot fundamentally improve NVS under large view changes, highlighting the need for more robust approaches and large-scale training. We have released our data to help advance self-driving and urban robotics simulation technology.
 
+<!-- ## 🗓️ News -->
+<!-- [2024.12.06] Our paper has been released on [arxiv]()! -->
 
 ## 🗓️ TODO
 - [ ] Data release