3D StreetUnveiler with Semantic-aware 2DGS - a simple baseline

ICLR 2025

Jingwei Xu1 · Yikai Wang2 · Yiqun Zhao3,6 · Yanwei Fu5 · Shenghua Gao3,4†
1 ShanghaiTech University 2 Nanyang Technological University 3 The University of Hong Kong 4 HKU Shanghai Intelligent Computing Research Center 5 Fudan University 6 Transcengram
(† denotes corresponding author)
arXiv Code
StreetUnveiler

StreetUnveiler: We achieve accurate reconstruction from in-car camera videos. With the aid of the proposed hard-label semantic 2D Gaussian Splatting and proposed time-reversal inpainting framework, we remove the unwanted objects with satisfactory appearance and geometry of occluded regions.

Demo

Main Experiment Comparison

Select which methods to compare.

(Please refresh if you don't see the video. Switch video will need some time to process.)

Training Data    VS    Ours



(Click here to switch the scene)

Novel View Synthesis

(Please refresh if you don't see the video. Switch video will need some time to process.)



(Click here to switch the scene)

BibTeX

 @inproceedings{xu2025streetunveiler,
      author       = {Jingwei Xu and Yikai Wang and Yiqun Zhao and Yanwei Fu and Shenghua Gao},
      title        = {3D StreetUnveiler with Semantic-aware 2DGS - a simple baseline},
      booktitle    = {The International Conference on Learning Representations (ICLR)},
      year         = {2025},
}