Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
97 tokens/sec
GPT-4o
53 tokens/sec
Gemini 2.5 Pro Pro
44 tokens/sec
o3 Pro
5 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

CondenseNet: An Efficient DenseNet using Learned Group Convolutions (1711.09224v2)

Published 25 Nov 2017 in cs.CV

Abstract: Deep neural networks are increasingly used on mobile devices, where computational resources are limited. In this paper we develop CondenseNet, a novel network architecture with unprecedented efficiency. It combines dense connectivity with a novel module called learned group convolution. The dense connectivity facilitates feature re-use in the network, whereas learned group convolutions remove connections between layers for which this feature re-use is superfluous. At test time, our model can be implemented using standard group convolutions, allowing for efficient computation in practice. Our experiments show that CondenseNets are far more efficient than state-of-the-art compact convolutional networks such as MobileNets and ShuffleNets.

User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (4)
  1. Gao Huang (179 papers)
  2. Shichen Liu (21 papers)
  3. Laurens van der Maaten (54 papers)
  4. Kilian Q. Weinberger (105 papers)
Citations (776)

Summary

  • The paper introduces learned group convolutions to prune DenseNet connections, reducing redundant computations.
  • It employs an efficient architecture that lowers parameter counts without sacrificing accuracy.
  • The method demonstrates significant benchmark improvements, offering practical benefits for resource-constrained applications.

A LaTeX Author Guidelines for CVPR Proceedings

This document, titled "LaTeX Author Guidelines for CVPR Proceedings" provides a comprehensive guide for authors intending to submit papers to the CVPR conference. The guide covers various aspects of paper preparation, from manuscript formatting to the incorporation of figures and references. Here, an expert overview is provided to synthesize the details and implications presented within this guide.

Key Guidelines and Recommendations

  1. Document Structure and Formatting:
    • Authors are required to follow a two-column format, ensuring that the text area is 6 7/8 inches wide by 8 7/8 inches high.
    • The main title should be placed 1 inch from the top edge of the first page, in 14-point boldface type, centered, with specific capitalization rules.
    • Author names and affiliations should be centered beneath the title in 12-point non-boldface type.
    • The body of the manuscript should use 10-point Times, single-spaced, justified both left and right, with indentation for paragraphs.
  2. Abstract and Sections:
    • The abstract should be fully justified, italicized, and in 12-point Times, positioned at the top of the left-hand column on the first page.
    • Papers should consist of appropriately numbered sections and subsections to facilitate easy navigation and citation.
  3. Figures and Tables:
    • Figures and tables should be placed within the two-column format and be numbered sequentially. Captions should be in a 9-point Roman type.
    • Authors should ensure that figures are legible when printed and resize fonts within figures to match the main text.
  4. Mathematics and Equations:
    • Displayed equations must be numbered for easy reference. The use of equation numbers instead of descriptors like "the equation second from the top of page 3 column 1" is encouraged.
    • Figure and table alignments with equations are discussed briefly, emphasizing the non-alternative nature of the provided ruler for final copy preparation.
  5. Blind Review and Citations:
    • Instructions for maintaining anonymity are crucial; authors should avoid terms like "my" or "our" when citing their own work.
    • Proper citation formats should be maintained, using square brackets for sequential citation numbering.
  6. Miscellaneous Recommendations:
    • Recommendations for avoiding common typographical errors, such as incorrect spacing after "e.g.", using the provided macros to manage citation styles, and keeping multiple citations in numerical order are noted.
    • Authors are advised to number all items, including figures, tables, and equations, for ease of reference.

Practical and Theoretical Implications

The document provides an essential framework that ensures consistency and readability across submissions, facilitating the review process and enhancing the presentation of research. By adhering to these guidelines, authors can minimize formatting discrepancies, which can detract from the perceived quality of their work.

Future Developments

While the guidelines cater to the current standards, future revisions might integrate dynamic elements such as interactive figures or embedded multimedia, reflecting the advancements in document processing and visualization technologies. Practically, these developments can foster richer paper presentations and engagement during the review and conference phases.

In summary, the provided guidelines serve as a meticulous roadmap for authors, promoting a standard of uniformity and precision essential for high-quality academic submissions to CVPR. The directives span technical formatting, manuscript structure, and best practices for maintaining anonymity during blind reviews, all aimed at streamlining the submission and review process.