Learning Human Motion Models for Long-term Predictions (1704.02827v2)

Abstract: We propose a new architecture for the learning of predictive spatio-temporal motion models from data alone. Our approach, dubbed the Dropout Autoencoder LSTM, is capable of synthesizing natural looking motion sequences over long time horizons without catastrophic drift or motion degradation. The model consists of two components, a 3-layer recurrent neural network to model temporal aspects and a novel auto-encoder that is trained to implicitly recover the spatial structure of the human skeleton via randomly removing information about joints during training time. This Dropout Autoencoder (D-AE) is then used to filter each predicted pose of the LSTM, reducing accumulation of error and hence drift over time. Furthermore, we propose new evaluation protocols to assess the quality of synthetic motion sequences even for which no ground truth data exists. The proposed protocols can be used to assess generated sequences of arbitrary length. Finally, we evaluate our proposed method on two of the largest motion-capture datasets available to date and show that our model outperforms the state-of-the-art on a variety of actions, including cyclic and acyclic motion, and that it can produce natural looking sequences over longer time horizons than previous methods.

• This document provides comprehensive LaTeX style and formatting guidelines for authors submitting manuscripts to the 3DV conference proceedings.
• It specifies requirements for manuscript length, two-column text layout, required fonts, and the proper formatting of figures, tables, and equations.
• The guide also details important submission rules, including adherence to anonymity for blind review, citation standards, and final submission protocol.

A Guide to Formatting 3DV Proceedings Submissions

The document titled "LaTeX Author Guidelines for 3DV Proceedings" serves as a comprehensive style guide for authors intending to submit their manuscripts to the 3DV conference series, focusing primarily on formatting and submission standards. Created to streamline the submission process and ensure uniform presentation across conference papers, this guide covers essential requirements in considerable detail.

Manuscript Preparation Guidelines

The paper outlines the expectations around manuscript preparation, emphasizing adherence to specified formatting standards. Authors must submit their manuscripts in English and ensure they are cognizant of the constraints around dual submissions. Importantly, 3DV restricts submissions that overlap significantly with other work under review, necessitating rigorous self-checks and declarations by the authors to prevent conflicts.

Formatting Specifications

A significant portion of the document is dedicated to formatting rules to be observed in the paper. This includes:

• Paper Length: Papers should not exceed eight pages, excluding references, and must adhere to specified margin settings. Overlength submissions are a basis for rejection.
• Text Layout: All submissions must be in a two-column format. Detailed specifications are provided for column width, margin settings, and header/footer requirements.
• Type-style and Fonts: Authors are instructed to utilize Times Roman or an equivalent font, with specific sizes and styles allocated to different sections such as titles, abstracts, and body text. This ensures readability and professional presentation.

Specific Considerations

Additional focus is given to items such as:

• Mathematics and Equations: Equations should be numbered consecutively to aid reference and review.
• Figures and Tables: Guidance is provided on the incorporation of graphical elements, underscoring their alignment and integration with the text to ensure consistency.
• References: The citation format and bibliography layout must comply with IEEE standards.

Anonymity and Blind Review

The guidelines stress the importance of maintaining author anonymity during the review process, particularly emphasizing how to cite previous work without revealing author identity. A clear explanation is provided to distinguish acceptable from unacceptable self-citations within double-blind review contexts.

Submission Instructions

The paper concludes with notes on submission protocol, including final copy requirements and miscellaneous advisories on elements such as footnotes, use of color, and the inclusion of illustrations. Moreover, authors are reminded to accompany their submissions with the IEEE copyright release form to facilitate the publication of their paper.

Implications and Speculations for Future Developments

While ostensibly a technical document, the guide reflects broader implications within the academic publication ecosystem—fostering standardization to minimize administrative overheads and enhancing the professional presentation. Enhanced consistency potentially sets the stage for improved automated processing of manuscripts and contributes to enduring archival legacies for academic literature.

Future iterations of such guidelines are likely to adapt to evolving technologies around digital publication and review processes. Increasing automation in formatting checks and the potential integration of collaborative tools in manuscript preparation may become focal points. Additionally, the burgeoning role of open-access platforms may prompt further refinements in the guidance provided to authors.

In essence, this paper's role as a formatting compass solidifies its importance in the scholarly landscape, ensuring that research is communicated clearly, accurately, and professionally.