CUS3D: A New Comprehensive Urban-Scale Semantic Segmentation 3D Benchmark Dataset

Lin Gao, Yu Liu, Shen Yan, Yuxiang Liu, Maojun Zhang

Keywords: Urban scene understanding, Comprehensive urban-scale dataset, Semantic segmentation, Benchmark dataset

Abstract: With the continuous advancement of smart city construction, the availability of large-scale and semantically enriched datasets is essential for enhancing the machine’s ability to understand urban scene. When dealing with large-scale scene, mesh data has a distinct advantage over point cloud data, as it can provide inherent geometric topology information and consume low memory space. However, existing publicly available large-scale scene mesh datasets have limitations in scale and semantic richness, and cannot cover a wider range of urban semantic information. Moreover, the prevailing large-scale 3D datasets mainly consist of a single data type, which restricts the wide applicability of benchmark applications and hinders the further development of 3D semantic segmentation techniques in urban scene. To address these issues, we propose a comprehensive urban-scale semantic segmentation benchmark dataset. This dataset provides finely annotated point cloud and mesh data types for 3D, as well as high-resolution original 2D images with detailed 2D semantic annotations. It is well suited for various research pursuits on semantic segmentation methodologies. The dataset covers a vast area of approximately 2.85 square kilometers, containing 10 semantic labels that span both urban and rural scenes. Each 3D point or triangular mesh in the dataset is meticulously labeled with one of ten semantic categories. We evaluate the performance of this novel benchmark dataset using 6 widely adopted deep learning baselines. The dataset will be publicly available upon the publish of the paper.

Primary Area: datasets and benchmarks

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 7650