Deep Geometric Prior for Surface Reconstruction (1811.10943v2)

Abstract: The reconstruction of a discrete surface from a point cloud is a fundamental geometry processing problem that has been studied for decades, with many methods developed. We propose the use of a deep neural network as a geometric prior for surface reconstruction. Specifically, we overfit a neural network representing a local chart parameterization to part of an input point cloud using the Wasserstein distance as a measure of approximation. By jointly fitting many such networks to overlapping parts of the point cloud, while enforcing a consistency condition, we compute a manifold atlas. By sampling this atlas, we can produce a dense reconstruction of the surface approximating the input cloud. The entire procedure does not require any training data or explicit regularization, yet, we show that it is able to perform remarkably well: not introducing typical overfitting artifacts, and approximating sharp features closely at the same time. We experimentally show that this geometric prior produces good results for both man-made objects containing sharp features and smoother organic objects, as well as noisy inputs. We compare our method with a number of well-known reconstruction methods on a standard surface reconstruction benchmark.

• The paper introduces a deep geometric prior that serves as an implicit regularizer for accurate surface reconstruction from point clouds.
• It leverages a neural network-based implicit representation to capture complex surface details without traditional parameterization.
• Experimental results show robust reconstruction of detailed surfaces even from noisy or sparse data, highlighting its practical applications.

Overview of "LaTeX Author Guidelines for CVPR Proceedings"

The document titled "LaTeX Author Guidelines for CVPR Proceedings" serves as a comprehensive style guide for authors submitting their manuscripts for the CVPR (Computer Vision and Pattern Recognition) conference. This paper provides explicit instructions on a variety of formatting and submission guidelines tailored to ensure uniformity and clarity in published conference proceedings. Such guidelines are essential in maintaining the standardization required for high-quality academic dissemination in the field of computer vision.

Core Elements and Structure

The document systematically details multiple aspects of paper preparation and submission, which include:

• Language and Dual Submission: Authors are mandated to draft their manuscripts exclusively in English. It also discusses the policies surrounding dual submission, cautioning against the simultaneous submission of identical work to multiple venues, except with justification and cross-reference protocols in place.
• Paper Length and Formatting: A strict maximum length of eight pages for the main paper content is enforced, with an unlimited page allowance for references. This is pivotal in maintaining the succinctness and focus of conference papers, while also providing opportunity for exhaustive bibliographical referencing.
• Review Protocols: Details are given on maintaining anonymity during blind review processes, helping to ensure impartiality in manuscript evaluations.
• Mathematics and Equations: It emphasizes the necessity of numbering sections and equations. This precision facilitates ease of reference and discussion by reviewers and readers, thereby enhancing the scientific dialogue.
• Type-style and Fonts: The document prescribes the use of Times Roman font, specifying sizes for different sections and ensuring consistency and readability across submissions. The two-column format is mandated, reflecting a common style in conference proceedings.
• Illustrations and Graphical Content: Authors are advised on the presentation of graphics, charts, and tables, with specific emphasis on resolution and clarity, recognizing the likelihood of printed review copies being utilized by evaluators.

Implications and Future Developments

The guidelines highlight the necessity of adhering to specific formatting standards to facilitate seamless integration into the larger compendium of CVPR proceedings. By enforcing these guidelines, CVPR ensures the accessibility, professionalism, and comparability of the presented research.

From a practical standpoint, the precision in formatting allows for automated typesetting and digital archiving processes to function optimally, reducing manual intervention and the inherent risk of error. Theoretically, this streamlined presentation protocol enhances the interpretive quality of the research, allowing peers to more readily focus on the scientific contributions without distraction from formatting inconsistencies.

In future developments, we can envision a continued evolution of these guidelines to encompass new typesetting technologies and presentation media. As scholarly communication increasingly embraces digital-native formats, it will be essential for such guidelines to incorporate interactive and multimedia elements, which can further engage audiences and enhance the communication of complex ideas.

In summary, the "LaTeX Author Guidelines for CVPR Proceedings" document serves as an essential blueprint for authors aiming to contribute to the CVPR conference, reinforcing the cultural and scientific standards that underpin successful academic publishing in computer vision research.