Synthesizing Images of Humans in Unseen Poses (1804.07739v1)

Abstract: We address the computational problem of novel human pose synthesis. Given an image of a person and a desired pose, we produce a depiction of that person in that pose, retaining the appearance of both the person and background. We present a modular generative neural network that synthesizes unseen poses using training pairs of images and poses taken from human action videos. Our network separates a scene into different body part and background layers, moves body parts to new locations and refines their appearances, and composites the new foreground with a hole-filled background. These subtasks, implemented with separate modules, are trained jointly using only a single target image as a supervised label. We use an adversarial discriminator to force our network to synthesize realistic details conditioned on pose. We demonstrate image synthesis results on three action classes: golf, yoga/workouts and tennis, and show that our method produces accurate results within action classes as well as across action classes. Given a sequence of desired poses, we also produce coherent videos of actions.

• The paper introduces a novel method that disentangles pose and appearance to synthesize realistic human images in new configurations.
• It employs convolutional networks and adversarial training techniques to ensure high-quality image generation even under occlusion and varied clothing conditions.
• Empirical results demonstrate significant improvements over baselines, underscoring its potential for future applications in graphics and virtual reality.

Overview of "Author Guidelines for CVPR Proceedings"

The document titled "Author Guidelines for CVPR Proceedings" serves as a comprehensive guide for authors preparing submissions for the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). It covers a myriad of formatting and stylistic norms that are crucial to adhere to in order to ensure compliance with the conference's submission standards.

Content Summary

The guidelines provide explicit instructions on various aspects of manuscript preparation, including paper length, language, style, and formatting requirements. The authors emphasize the importance of adhering to the prescribed document structure, which facilitates consistency and facilitates the review process.

1. Formatting and Style: The document stipulates that papers must be presented in a two-column format with specific requirements for margins, font size, and page numbering. Authors must use Times or Times Roman for all text, ensuring the uniformity across submissions.
2. Blind Review Policy: A significant portion of the document is dedicated to detailing the blind review process. It clarifies the differences between citing one's previous work and anonymizing a manuscript for review, highlighting practices that can inadvertently reveal authorship.
3. Document Length: The guidelines set a firm limit on the length of submissions, permitting a maximum of eight pages for the main text, while reference sections do not contribute to the page count. This limit necessitates precise and concise articulation of research contributions.
4. Figures and Tables: The paper includes specific instructions on the inclusion and formatting of graphical elements, advising authors to ensure that visuals are legible and adequately sized for print.
5. Mathematical Expressions: There are explicit instructions to number all sections and equations comprehensively. This enhances the manuscript's navigability, allowing future readers to reference specific equations easily.
6. References and Citations: The guidelines dictate the expected style for bibliographic entries and in-text citations. References are to be listed in numerical order, with priority given to proper formatting to enable straightforward retrieval and verification of sources.

Implications and Speculative Developments

The document primarily serves a functional role in ensuring the standardization of submissions for CVPR, and its implications are largely procedural. By enforcing a consistent format, the guidelines aim to facilitate efficient review and dissemination of research findings.

From a meta perspective, the implications of such structured guidelines are far-reaching. They ensure that works are more readily comparable, streamlining the evaluation process for reviewers and aiding the indexing and archival processes. As AI-related conferences continue to evolve, the guideline structure might witness adaptations in response to emerging practices around collaborative authorship, dataset sharing, and code repositories.

The emphasis on meticulous formatting and the blind review process indicates a continual move towards transparency and fairness in peer review. This might inspire further research and development of automated tools to assist researchers in adhering to such guidelines, perhaps integrating AI to automate checks for compliance before submission.

Conclusion

This document establishes a thorough foundation for authors submitting to CVPR, covering all pertinent aspects of manuscript preparation. Its role is not only in maintaining the professional standards of the conference but also in influencing how research is communicated within the computer vision community. As the field progresses, these guidelines may integrate more adaptive elements reflecting broader shifts in digital manuscript preparation and submission practices.