Geometric Deep Learning for Computer Vision and Image Analysis: A Survey of Recent Advances and Future Directions
Saved in:
| Title: | Geometric Deep Learning for Computer Vision and Image Analysis: A Survey of Recent Advances and Future Directions |
|---|---|
| Authors: | Suresh A. J. |
| Source: | International Journal of Intelligent Systems and Applications in Engineering; Vol. 12 No. 1 (2024); 847 – 853 |
| Publisher Information: | International Journal of Intelligent Systems and Applications in Engineering, 2023. |
| Publication Year: | 2023 |
| Subject Terms: | Geometric Deep Learning, Computer Vision, Image Analysis, Graph Neural Networks, 3D Shape Analysis, Spectral Methods, Message-Passing Algorithms |
| Description: | Geometric Deep Learning (GDL) has emerged as a powerful framework for addressing complex computer vision and image analysis tasks by extending traditional deep learning techniques to non-Euclidean data structures such as graphs, manifolds, and meshes. This survey provides a comprehensive overview of recent advances in GDL for computer vision, highlighting its application in areas such as 3D shape analysis, medical imaging, scene understanding, and object recognition. We discuss key architectural innovations, including graph neural networks, spectral methods, and message-passing algorithms, that enable the effective representation and processing of geometric data. Furthermore, we explore challenges such as computational complexity and generalization across diverse domains. Lastly, we outline potential future research directions, including the integration of GDL with multimodal learning, improved scalability, and the development of more robust and interpretable models. This survey emphasizes GDL’s growing significance in advancing state-of-the-art computer vision techniques and its potential to solve increasingly complex tasks. |
| Document Type: | Article |
| File Description: | application/pdf |
| Language: | English |
| ISSN: | 2147-6799 |
| Access URL: | https://www.ijisae.org/index.php/IJISAE/article/view/7035 |
| Rights: | CC BY SA |
| Accession Number: | edsair.issn21476799..beeabe9ff1f1de7b96daf7257bb0e5d1 |
| Database: | OpenAIRE |
Nájsť tento článok vo Web of Science | Abstract: | Geometric Deep Learning (GDL) has emerged as a powerful framework for addressing complex computer vision and image analysis tasks by extending traditional deep learning techniques to non-Euclidean data structures such as graphs, manifolds, and meshes. This survey provides a comprehensive overview of recent advances in GDL for computer vision, highlighting its application in areas such as 3D shape analysis, medical imaging, scene understanding, and object recognition. We discuss key architectural innovations, including graph neural networks, spectral methods, and message-passing algorithms, that enable the effective representation and processing of geometric data. Furthermore, we explore challenges such as computational complexity and generalization across diverse domains. Lastly, we outline potential future research directions, including the integration of GDL with multimodal learning, improved scalability, and the development of more robust and interpretable models. This survey emphasizes GDL’s growing significance in advancing state-of-the-art computer vision techniques and its potential to solve increasingly complex tasks. |
|---|---|
| ISSN: | 21476799 |