NeRF-RPN: A general framework for object detection in NeRFs (2211.11646v3)

Abstract: This paper presents the first significant object detection framework, NeRF-RPN, which directly operates on NeRF. Given a pre-trained NeRF model, NeRF-RPN aims to detect all bounding boxes of objects in a scene. By exploiting a novel voxel representation that incorporates multi-scale 3D neural volumetric features, we demonstrate it is possible to regress the 3D bounding boxes of objects in NeRF directly without rendering the NeRF at any viewpoint. NeRF-RPN is a general framework and can be applied to detect objects without class labels. We experimented NeRF-RPN with various backbone architectures, RPN head designs and loss functions. All of them can be trained in an end-to-end manner to estimate high quality 3D bounding boxes. To facilitate future research in object detection for NeRF, we built a new benchmark dataset which consists of both synthetic and real-world data with careful labeling and clean up. Code and dataset are available at https://github.com/lyclyc52/NeRF_RPN.

Author Guidelines for CVPR Proceedings

This document presents a comprehensive set of guidelines designed to aid authors in preparing their manuscripts for conference proceedings at the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). By systematically detailing the structural and formatting requirements, this guide plays an essential role in standardizing submissions, thereby facilitating an efficient and cohesive review process.

Manuscript Structure and Length

The guidelines emphasize that manuscripts, excluding references, must not exceed eight pages. The references section is required yet does not contribute to the page count, offering flexibility with length coverage for bibliographic entries. For example, a submission might encompass eight pages of content plus two additional pages devoted to references. Notably, CVPR no longer imposes extra charges for exceeding page limits in submissions, potentially easing financial considerations for authors. However, adherence to the page limit is crucial, as non-compliant submissions will not be subjected to review.

Blind Review Protocol

A critical component of the submission process pertains to maintaining anonymity for the purposes of double-blind review. The guidelines clarify misconception points regarding this requirement. Specifically, the document makes clear that authors should avoid personal pronouns when citing their own previous work, instead opting for third-person references. This approach seeks to eliminate any potential identifiers that could bias the review process. Authors are reminded to take meticulous care in ensuring that anonymity is preserved throughout the manuscript.

Practical Formatting Requirements

This guide provides explicit instructions on various practical aspects of document formatting:

• Page Layout: Authors are instructed to employ a two-column layout with specific measurements for columns and margins, including precise spacing for headers, footers, and page numbers.
• Font and Typography: Recommendations are made for using specific fonts (e.g., Times Roman) and font sizes throughout various manuscript sections, including titles, headings, and body text.
• Figures, Tables, and Equations: Detailed advice is given on optimizing the presentation of figures and tables, including size instructions and centroid alignment. Authors are encouraged to use LaTeX commands for numbering and referencing in a manner that supports clear navigation within the manuscript.
• Color Usage: Considerations around color are addressed, with explicit advice to avoid reliance solely on color for conveying information, accommodating those with color vision deficiencies.

Additional Considerations

The manuscript preparation advice extends into numerous nuanced aspects of writing and presentation:

• Ruler and Paper ID: Including a printed ruler and ensuring the visibility of the Paper ID is encouraged for all submitted drafts to enhance the review experience.
• Cross-references and Citations: Authors are urged to use systematic cross-referencing tools, which not only enhance readability but ensure coherence in supporting cited work.
• Mathematics and Footnotes: Emphasis is placed on numbering all displayed equations and the sparing use of footnotes to maintain the focus on the main text flow.

Broader Implications

This document serves as a valuable resource for authors, providing structured guidance that ensures clarity, uniformity, and professionalism in manuscript submissions to a leading computer vision conference. By establishing a standardized approach, CVPR aims to streamline the reviewing process, thereby contributing to the dissemination of well-presented research findings. The detailed protocol not only aids authors in technical matters but also aligns them with a wider framework for scientific communication within the field. All participants in the process benefit: authors submit polished work, reviewers access clearly organized submissions, and the broader community gains from the increased quality and reliability of published research.

In conclusion, while purely instructive in nature, this document's comprehensive scope and meticulous presentation work to uphold the integrity and efficiency of the peer-review process, enhancing the impact and reach of research disseminated through the CVPR platform. Future developments might see these guidelines adapted further as technological and publishing landscapes evolve, continuing to support the advancement of computer science research.