Call for Papers » Deep Learning & Image Processing

Deep Learning & Image Processing Conference - Call for Papers

We invite your original quality submissions to the Computer Vision Conference 2027 Track on Deep Learning & Image Processing.

This track covers the core algorithmic and architectural foundations of computer vision, from convolutional and self-supervised learning methods to the architecture design decisions that determine how well a vision model actually performs. We invite contributions on object detection, feature detection, segmentation, and scene understanding, as well as three-dimensional vision and reconstruction work that recovers structure from images rather than treating them as flat data. Submissions may include architecture papers proposing new building blocks or training strategies, explainability work that opens up what a vision model has actually learned, or empirical studies benchmarking established approaches under more realistic conditions than their original evaluation. Evidence that a method generalizes across datasets, domains, or viewpoints, rather than only the benchmark it was tuned on, is particularly valued, along with genuine efforts at interpretability rather than post-hoc visualization alone. The track ultimately looks for advances in how vision systems perceive and represent the world, at whatever level of the stack that progress happens.

Round 1 (Closed)

Submission Deadline: 01 September 2026
Notification of Acceptance: 15 September 2026
Camera-Ready Submission: 01 October 2026
Conference Dates: 15-16 April 2027

Round 2 (Open)

Submission Deadline: 01 October 2026
Notification of Acceptance: 01 November 2026
Camera-Ready Submission: 01 December 2026
Conference Dates: 15-16 April 2027

All times are in Anywhere on Earth (AoE) time zone.

Submit Now

Topics of Interest

Convolutional Neural Networks: Foundational deep learning architecture for spatial feature extraction, underlying much of modern computer vision.
Self-supervised Learning: Training methods that learn visual representations from unlabeled data using pretext tasks, reducing dependence on manual annotation.
Explainable AI (XAI): Techniques for making model predictions interpretable to end users and stakeholders, including feature attribution, surrogate models, and human-centered explanation design.
Neural Network Architecture Design: Principled and automated approaches to designing neural network architectures for vision tasks, including neural architecture search and efficiency-oriented design for deployment-constrained settings.
Object Detection: Algorithms for localizing and classifying objects within images or video frames.
Feature Detection: Identification of distinctive points, edges, or regions in images that support matching, tracking, and recognition tasks.
Image Segmentation: Pixel-level partitioning of images into semantically meaningful regions, spanning semantic, instance, and panoptic segmentation and their application to downstream recognition tasks.
Visual Recognition: Classification and identification of objects, scenes, and activities from visual data.
Scene Understanding: Holistic interpretation of visual scenes, including object relationships, layout, and context.
3D Vision & Reconstruction: Recovering three-dimensional structure from single or multiple images, including depth estimation, point cloud processing, and neural implicit representations such as NeRF-style scene reconstruction.

Submission Guidelines

We accept initial submission in PDF format only.
To adhere to the double-blind peer review process, authors must refrain from including any details related to their identity, such as author names, affiliations, or country information, within the paper.
Authors are responsible for ensuring that their paper is thoroughly checked and proofread prior to submission. Please note that no changes can be made to the paper once it has been submitted for review. However, if the paper is accepted, minor revisions may be made before final publication to address any areas for improvement.
When submitting the camera-ready version of the paper, please ensure that all author information is included, as required by the paper formatting guidelines. Author names, email addresses, and their sequence will be considered final as per the submitted camera-ready version.
Submissions must primarily consist of original work. No more than 25% of the content should be derived from previously published material by the same authors, and any such reused content must be properly cited. Therefore, at least 75% of the manuscript should contain new, unpublished material.

Downloads

Formatting Guidelines

Manuscripts should typically be confined to 18 pages for the main text (excluding references and appendices of upto 7 pages). Submissions that exceed this limit may be considered at the discretion of the conference chair, particularly for review or survey papers.
Manuscripts should present original research that aligns with the conference themes and topics.
It is the responsibility of the authors to ensure the accuracy of all citations, quotations, figures, maps, and tables included in their manuscript.
The abstract should adhere to the 150-250 word limit as per the conference guidelines.
Figures and tables should be placed within the manuscript at the appropriate locations where they are referenced and must be clear and legible.
All figures, tables, references, appendices, algorithms, annexures, and supplementary materials must be clearly numbered, and properly cited in the text.
Tables must be neatly aligned within the page margins and must not overflow the page boundaries.
Figure captions and table headings should be concise, unique, and ideally limited to a maximum of two lines.
References should be formatted correctly and consistently, including all the necessary details such as author names, paper titles, publication sources, and years. Every reference listed must be cited in the main text.
The use of foreign languages in the manuscript is not allowed unless accompanied by a corresponding English translation.
The manuscript must comply with the formatting guidelines specified in the provided template.

Review Process

Submissions will undergo an anonymous review process (double-blind). As a result, authors must ensure their manuscript does not disclose their identity or affiliation. Specifically, authors should remove their names and affiliations from the manuscript, anonymize any citations or references to their own related work, and omit any acknowledgments, including funding sources.

Each paper will be evaluated by a minimum of three regular program committee members or two senior program committee members. Acceptance will be determined based on the originality, technical depth, elegance, impact (either practical or theoretical), and overall presentation quality.

Original: The paper should introduce a new idea, project, or issue; provide fresh insights into existing research; present new findings; or offer a novel perspective on established information.
Engaging: The presentation should be interactive, engaging the audience, or address community needs in a way that is likely to attract substantial conference attendance.
Significant: The paper should tackle critical issues related to improving the effectiveness of current methods, and its content should be clear and broadly applicable.
Quality: Any claims made in the paper must be supported by sufficient data, properly reference existing work, and honestly acknowledge any limitations.
Clear: The outcomes of the paper should be presented in a manner that is easily understood.
Relevant: The paper’s subject matter must align with at least one of the conference’s designated tracks.

Use of AI-Assisting Tools

The conference permits the responsible use of Generative AI (GenAI) tools in manuscript preparation. Authors are required to observe the following:

Declare Usage – Any use of GenAI must be disclosed in the manuscript (e.g., in a Declaration on Generative AI section).
Maintain Responsibility – Authors must critically review and edit AI-assisted text; unedited AI output is not acceptable.
No AI Authorship – GenAI tools cannot be listed as authors. Authorship is limited to humans who take full responsibility for the work.
Limit Role of AI – GenAI may support drafting and language refinement, but the development of scientific ideas, arguments, results, and conclusions must be carried out by human authors.
Accountability – Authors remain fully accountable for the accuracy, originality, and integrity of their manuscripts.

Note: Non-compliance with these guidelines will be treated as academic misconduct and may result in the removal of published work.

Other Tracks at Computer Vision Conference 2027

View the full Call for Papers for Computer Vision Conference 2027 »