Depth map estimation method based on deformable convolutional neural network

Zihao Yin; Shan Leng

doi:10.1117/12.3056041

9 January 2025 Depth map estimation method based on deformable convolutional neural network

Zihao Yin, Shan Leng

Author Affiliations +

Proceedings Volume 13486, Fourth International Conference on Computer Vision, Application, and Algorithm (CVAA 2024); 134860P (2025) https://doi.org/10.1117/12.3056041
Event: Fourth International Conference on Computer Vision, Application, and Algorithm (CVAA 2024), 2024, Chengdu, China

Abstract

This paper proposes a depth map estimation method based on Deformable Convolutional Neural Networks (DCNNs). Traditional single image depth estimation methods often struggle to accurately capture irregular shapes and details in complex scenes, resulting in suboptimal spatial resolution and object boundary reconstruction. To address these challenges, we introduce deformable convolution modules that enable convolutional kernels to adaptively adjust their sampling positions, thereby enhancing the model's capability to handle complex geometric structures and dynamic deformations. Deformable convolution features high adaptability, superior handling of complex scenes, and enhanced boundary clarity. The position offsets of convolutional kernels are adaptively adjusted through learning, allowing the model to flexibly handle features of various shapes and scales, thus excelling in capturing irregular shapes and complex details. In scenes with abundant details and deformations, deformable convolution can extract features more accurately, significantly improving depth estimation accuracy. Additionally, by finely adjusting sampling points, deformable convolution effectively reduces blurring and distortion at depth map boundaries, excelling in reconstructing small objects and complex edges. Experimental results demonstrate that the proposed method significantly outperforms existing approaches in terms of depth estimation accuracy and boundary clarity, particularly excelling in complex scenes and small object reconstruction. Our research highlights the extensive application potential of deformable convolution in depth map estimation, providing robust support for future computer vision tasks.

(2025) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Zihao Yin and Shan Leng "Depth map estimation method based on deformable convolutional neural network", Proc. SPIE 13486, Fourth International Conference on Computer Vision, Application, and Algorithm (CVAA 2024), 134860P (9 January 2025); https://doi.org/10.1117/12.3056041

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

;

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available

Members: $17.00

Non-members: $21.00 ADD TO CART

PROCEEDINGS
9 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Convolution

Deformation

Depth maps

Computer vision technology

Convolutional neural networks

Feature extraction

Edge detection

Show All Keywords

Keywords/Phrases

Search In:

Publication Years