ViewAL: Active Learning with Viewpoint Entropy for Semantic Segmentation (1911.11789v2)

Abstract: We propose ViewAL, a novel active learning strategy for semantic segmentation that exploits viewpoint consistency in multi-view datasets. Our core idea is that inconsistencies in model predictions across viewpoints provide a very reliable measure of uncertainty and encourage the model to perform well irrespective of the viewpoint under which objects are observed. To incorporate this uncertainty measure, we introduce a new viewpoint entropy formulation, which is the basis of our active learning strategy. In addition, we propose uncertainty computations on a superpixel level, which exploits inherently localized signal in the segmentation task, directly lowering the annotation costs. This combination of viewpoint entropy and the use of superpixels allows to efficiently select samples that are highly informative for improving the network. We demonstrate that our proposed active learning strategy not only yields the best-performing models for the same amount of required labeled data, but also significantly reduces labeling effort. For instance, our method achieves 95% of maximum achievable network performance using only 7%, 17%, and 24% labeled data on SceneNet-RGBD, ScanNet, and Matterport3D, respectively. On these datasets, the best state-of-the-art method achieves the same performance with 14%, 27% and 33% labeled data. Finally, we demonstrate that labeling using superpixels yields the same quality of ground-truth compared to labeling whole images, but requires 25% less time.

• The paper provides comprehensive instructions for authors on preparing manuscripts for submission to the CVPR conference using the LaTeX document system.
• It details specific formatting requirements including abstract style, maximum paper length (8 pages excluding references), figure placement, and equation numbering.
• Following these guidelines ensures submission uniformity, facilitates the blind review process, and helps maintain high-quality standards for published research.
• Adhering to these guidelines ensures submission uniformity, facilitates the blind review process, and helps maintain high-quality standards for published research.

Guidelines for Formatting CVPR Proceedings Using \LaTeX

The paper "Author Guidelines for CVPR Proceedings" provides comprehensive instructions for preparing a manuscript for submission to the Conference on Computer Vision and Pattern Recognition (CVPR) using the \LaTeX document preparation system. This document is crucial for authors aiming to conform to the standards set forth by the IEEE Computer Society and ensures that submissions are consistent in formatting for review and publication.

Key Components and Instructions

1. Abstract and Introduction: The abstract must be italicized, fully justified, and placed at the top left-hand column, with specific font size and type requirements. The introduction sets the stage for adhering to these guidelines, emphasizing modifications and essential practices that authors must follow.
2. Language and Submissions: Manuscripts must be written in English, and dual submission policies should be referred to on the CVPR conference page.
3. Paper Length and Format:
   • The primary text, excluding references, should not exceed eight pages.
   • Overlength papers are not considered for review.
   • There are no additional charges for extra pages beyond the reference sections.
   • Authors are instructed on maintaining specific font sizes and spacing requirements for different segments of the paper such as headings, footnotes, and tables.
4. Blind Review Requirements: Authors are guided on anonymizing submissions until the blind review phase is complete, with clear directives on citations and self-reference.
5. Math and Equations: All sections and displayed equations must be numerically labeled, ensuring ease of reference and discussion within the manuscript.
6. Figures and Illustrations: Detailed instructions are provided for the placement, captioning, and formatting of figures to ensure that they align well with the overall text.
7. Type-Style and Fonts: The paper specifies font types and sizes for different parts of the manuscript to maintain uniformity and readability across all submissions.

Practical and Theoretical Implications

This guide addresses practical challenges in manuscript preparation, delineating a structured pathway for authors to follow to avoid common pitfalls in the submission process. By standardizing formats, these guidelines endeavor to streamline the review process, allowing reviewers to focus on the content rather than inconsistencies in formatting.

Theoretically, such detailed guidelines support the goal of producing a cohesive body of work across diverse submissions, facilitating better collective understanding and comparison among research contributions from the computer vision community.

Future Developments

As academic conferences evolve with technological and academic advancements, we can anticipate future updates to these guidelines. These changes may include incorporation of more advanced manuscript preparation tools or changes in the submission portal to enhance ease of use for authors and reviewers alike. Furthermore, developments in typesetting technologies and digital publication standards may influence how guidelines are adapted in the future.

In conclusion, the "Author Guidelines for CVPR Proceedings" paper is an indispensable reference for authors submitting to CVPR, ensuring that all manuscripts meet rigorous scholarly standards that support efficient review and high-quality publication.