Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
102 tokens/sec
GPT-4o
59 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
6 tokens/sec
GPT-4.1 Pro
50 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Image Super-Resolution via Iterative Refinement (2104.07636v2)

Published 15 Apr 2021 in eess.IV, cs.CV, and cs.LG

Abstract: We present SR3, an approach to image Super-Resolution via Repeated Refinement. SR3 adapts denoising diffusion probabilistic models to conditional image generation and performs super-resolution through a stochastic denoising process. Inference starts with pure Gaussian noise and iteratively refines the noisy output using a U-Net model trained on denoising at various noise levels. SR3 exhibits strong performance on super-resolution tasks at different magnification factors, on faces and natural images. We conduct human evaluation on a standard 8X face super-resolution task on CelebA-HQ, comparing with SOTA GAN methods. SR3 achieves a fool rate close to 50%, suggesting photo-realistic outputs, while GANs do not exceed a fool rate of 34%. We further show the effectiveness of SR3 in cascaded image generation, where generative models are chained with super-resolution models, yielding a competitive FID score of 11.3 on ImageNet.

Overview of "LaTeX Author Guidelines for ICCV Proceedings"

The paper "LaTeX Author Guidelines for ICCV Proceedings" provides comprehensive instructions for authors preparing their manuscripts for submission to the International Conference on Computer Vision (ICCV). These guidelines are pivotal for ensuring consistency and clarity in the presentation of research papers, facilitating the peer review process and subsequent publication in conference proceedings.

Abstract and Introduction

The abstract section, formatted uniquely in fully-justified italicized text, sets the tone for the rest of the document. Immediately following are the initial instructions and changes since previous iterations, indicating a continuous refinement to suit evolving submission requirements.

Manuscript Preparation

The paper explores several critical aspects of manuscript preparation, including:

  • Language: English is mandatory.
  • Dual Submission: Authors are reminded to refer to the official ICCV guidelines regarding dual submissions.
  • Paper Length: Strict limitations are imposed, with a maximum of eight pages excluding references. A clear stance against overlength papers is emphasized: such submissions will not be reviewed. Authors must adhere strictly to this rule without manipulating margins or formatting, as the review process cannot fairly assess papers exceeding the stipulated length.
  • Blind Review Process: The paper clarifies common misconceptions regarding the anonymizing process. It emphasizes the importance of avoiding terms that could reveal the authors' identities and provides examples of acceptable and unacceptable practices.

Technical Specifications

The Ruler: The inclusion of a printed ruler in the submission version is intended to aid reviewers in referencing specific lines, although it should be omitted in the final camera-ready version.

Mathematics: The document advises thorough numbering of all sections and displayed equations, supporting easy reference for readers and reviewers.

Illustrations and Figures: Authors are instructed on the placement and captioning of figures and tables. Suggestions include keeping graphics centered, matching font sizes within figures to the body text, and using \verb+\includegraphics+ for figure embedding. Examples illustrate proper caption formatting, ensuring readability and clarity both electronically and in print.

Formatting Details

Margins and Page Numbering: Specific dimensions for text and illustrations ensure uniformity. For review submissions, page numbers should appear in the footer, centered, and commencing from page one. However, these should be omitted in the final copy submitted for publication, as the publisher will handle this aspect.

Type-styles and Fonts: Detailing font requirements, the guide stipulates using Times Roman or its closest available alternative. Different point sizes for various sections and headings ensure hierarchical clarity.

Author Attributions

References and Footnotes: References must be listed and numbered in a 9-point Times font, meticulously formatted for ease of cross-referencing within the text. The paper discourages excessive use of footnotes, suggesting inline incorporation of supplementary information to maintain narrative flow.

Ethical and Submission Standards

Acknowledgements and Citing Open Challenge Results: Adhering to the double-blind review policy, acknowledgements are reserved for the final copy, and authors must anonymize their participation in open challenges when presenting their results.

Practical Implications and Future Considerations

The guidelines presented address a fundamental need for standardization in manuscript preparation. By ensuring all submissions conform to a uniform format, the ICCV facilitates a smoother review process, enhancing the efficiency and fairness of academic evaluations. These detailed instructions also equip authors with the knowledge to prevent common submission pitfalls, thereby increasing their chances of acceptance.

Future developments in AI and paper submission guidelines will likely continue to evolve in parallel, reflecting advances in document preparation technologies. The ongoing refinement of these guidelines is essential to meet the changing landscape of academic research dissemination.

In conclusion, the "LaTeX Author Guidelines for ICCV Proceedings" serves as an indispensable resource for researchers aiming to contribute to ICCV. Its detailed stipulations ensure that submitted manuscripts are polished, professional, and conform to the high standards expected by one of the premier conferences in computer vision.

User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (6)
  1. Chitwan Saharia (16 papers)
  2. Jonathan Ho (27 papers)
  3. William Chan (54 papers)
  4. Tim Salimans (46 papers)
  5. David J. Fleet (47 papers)
  6. Mohammad Norouzi (81 papers)
Citations (1,490)
Youtube Logo Streamline Icon: https://streamlinehq.com