Towards Semantic Segmentation of Urban-Scale 3D Point Clouds: A Dataset, Benchmarks and Challenges (2009.03137v3)

Abstract: An essential prerequisite for unleashing the potential of supervised deep learning algorithms in the area of 3D scene understanding is the availability of large-scale and richly annotated datasets. However, publicly available datasets are either in relative small spatial scales or have limited semantic annotations due to the expensive cost of data acquisition and data annotation, which severely limits the development of fine-grained semantic understanding in the context of 3D point clouds. In this paper, we present an urban-scale photogrammetric point cloud dataset with nearly three billion richly annotated points, which is three times the number of labeled points than the existing largest photogrammetric point cloud dataset. Our dataset consists of large areas from three UK cities, covering about 7.6 km^2 of the city landscape. In the dataset, each 3D point is labeled as one of 13 semantic classes. We extensively evaluate the performance of state-of-the-art algorithms on our dataset and provide a comprehensive analysis of the results. In particular, we identify several key challenges towards urban-scale point cloud understanding. The dataset is available at https://github.com/QingyongHu/SensatUrban.

• The paper outlines the CVPR author response process by detailing a strict one-page, two-column rebuttal format for addressing reviewer feedback.
• It specifies essential LaTeX formatting rules, including figure/table placement, typography, and systematic enumeration of sections and references.
• The guidelines emphasize clarifying content and correcting inaccuracies without introducing novel contributions, maintaining focus on the original submission.

LaTeX Guidelines for Author Response in CVPR

The document "LaTeX Guidelines for Author Response" pertains specifically to the Computer Vision and Pattern Recognition (CVPR) conference, offering concise directives on formulating an author rebuttal after the initial paper reviews are received. This process allows authors to address review feedback, enhancing the communication of research submissions.

Author Rebuttal Purpose and Constraints

The primary objective of the author rebuttal is to address factual inaccuracies or to provide supplementary information that reviewers have sought. Importantly, it is not intended for introducing novel contributions such as theorems, algorithms, or experimental results that were not part of the original submission. This keeps the focus on clarifying existing content and correcting misunderstandings, not expanding the scope of the submission.

A significant policy established in 2018 by the PAMI Technical Committee discourages reviewers from requesting additional experiments during the rebuttal phase. Authors are advised against incorporating new experimental results, and any such inclusions should be disregarded by reviewers in their final evaluations. The permissible content includes figures or tables that illustrate comparisons or reported results relevant to the discussion.

Formatting and Structure

The rebuttal must be a one-page PDF, adhering to the blind-submission protocol similar to the original submission. The document mandates a two-column format, with detailed specifications on dimensions and layout to maintain consistency and readability. Authors are reminded to update pertinent details such as the paper title and ID in the LaTeX file.

Sections, equations, and bibliographical references must be systematically enumerated. This is critical for facilitating precise referencing within the limited page space, ensuring clarity and ease of navigation through the content.

Graphical Content and Typography

For graphical inclusions, alignment is centrally required, and care must be taken to maintain print legibility. Resizing font and choosing appropriate line widths are emphasized to ensure visual clarity. Authors are advised to use the \includegraphics command effectively for sizing figures relative to the linewidth, preserving document integrity in various formats.

Figures and tables must have captions in 9-point Roman type, enabling the seamless incorporation of mathematical content without aesthetic discord. This typographic detail emphasizes the importance of maintaining consistent visual standards across the rebuttal document.

Implications and Future Considerations

The structured approach to author responses promulgated by these guidelines ensures a balanced and focused discourse between authors and reviewers. The repercussions of such a protocol are evident in its capacity to streamline communication, uphold rigorous standards, and facilitate a clear understanding of submitted work.

As the processes around AI conferences continue to evolve, such guidelines could set a precedent for similar academic venues, possibly leading to further refinements in handling author-reviewer interactions. Continued adherence and adaptation to these directives will be essential in maintaining and enhancing the quality of scholarly discourse within the AI research community.