| 6. | Dvořák J; Hácha F; Arvanitis G; Podgorelec D; Moustakas K; Váša L Survey of Inter-Prediction Methods for Time-Varying Mesh Compression Journal Article In: Computer Graphics Forum, pp. e15278, 2025. @article{dvorak2025survey,
title = {Survey of Inter-Prediction Methods for Time-Varying Mesh Compression},
author = {Jan Dvořák and Filip Hácha and Gerasimos Arvanitis and David Podgorelec and Konstantinos Moustakas and Libor Váša},
url = {https://www.vvr.ece.upatras.gr/wp-content/uploads/sites/5/2025/01/Survey-of-Inter‐Prediction-Methods-for-Time‐Varying-Mesh-Compression_compressed.pdf, pdf/preprint},
doi = {https://doi.org/10.1111/cgf.15278},
year = {2025},
date = {2025-01-13},
urldate = {2025-01-13},
journal = {Computer Graphics Forum},
pages = {e15278},
abstract = {Abstract Time-varying meshes (TVMs), that is mesh sequences with varying connectivity, are a greatly versatile representation of shapes evolving in time, as they allow a surface topology to change or details to appear or disappear at any time during the sequence. This, however, comes at the cost of large storage size. Since 2003, there have been attempts to compress such data efficiently. While the problem may seem trivial at first sight, considering the strong temporal coherence of shapes represented by the individual frames, it turns out that the varying connectivity and the absence of implicit correspondence information that stems from it makes it rather difficult to exploit the redundancies present in the data. Therefore, efficient and general TVM compression is still considered an open problem. We describe and categorize existing approaches while pointing out the current challenges in the field and hint at some related techniques that might be helpful in addressing them. We also provide an overview of the reported performance of the discussed methods and a list of datasets that are publicly available for experiments. Finally, we also discuss potential future trends in the field.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Abstract Time-varying meshes (TVMs), that is mesh sequences with varying connectivity, are a greatly versatile representation of shapes evolving in time, as they allow a surface topology to change or details to appear or disappear at any time during the sequence. This, however, comes at the cost of large storage size. Since 2003, there have been attempts to compress such data efficiently. While the problem may seem trivial at first sight, considering the strong temporal coherence of shapes represented by the individual frames, it turns out that the varying connectivity and the absence of implicit correspondence information that stems from it makes it rather difficult to exploit the redundancies present in the data. Therefore, efficient and general TVM compression is still considered an open problem. We describe and categorize existing approaches while pointing out the current challenges in the field and hint at some related techniques that might be helpful in addressing them. We also provide an overview of the reported performance of the discussed methods and a list of datasets that are publicly available for experiments. Finally, we also discuss potential future trends in the field. |
| 5. | Vlachos C; Moustakas K High–Fidelity Haptic Rendering Through Implicit Neural Force Representation Proceedings Article In: International Conference on Human Haptic Sensing and Touch Enabled Computer Applications, pp. 493–506, Springer 2024. @inproceedings{vlachos2024high,
title = {High–Fidelity Haptic Rendering Through Implicit Neural Force Representation},
author = {Christoforos Vlachos and Konstantinos Moustakas},
url = {https://www.vvr.ece.upatras.gr/wp-content/uploads/sites/5/2024/11/high_fidelity_haptic_rendering.pdf, pdf/preprint},
doi = {10.1007/978-3-031-70058-3_40},
year = {2024},
date = {2024-11-03},
urldate = {2024-11-03},
booktitle = {International Conference on Human Haptic Sensing and Touch Enabled Computer Applications},
pages = {493–506},
organization = {Springer},
abstract = {The use of neural networks with periodic nonlinearities has been explored for the implicit representation and reconstruction of continuous-time signals. Building upon a previously published network for representing the Signed Distance Function (SDF) of a mesh surface, the Unit Normal Function (UNF) has been introduced. With both functions represented, a penalty-based haptic rendering method has been developed. The method performs well with very large meshes, outperforming other methods by generating continuous, high-fidelity forces free of discontinuities. It achieves high spatial accuracy by sampling a continuous implicit force function, enhancing the realism of haptic feedback in virtual environments.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The use of neural networks with periodic nonlinearities has been explored for the implicit representation and reconstruction of continuous-time signals. Building upon a previously published network for representing the Signed Distance Function (SDF) of a mesh surface, the Unit Normal Function (UNF) has been introduced. With both functions represented, a penalty-based haptic rendering method has been developed. The method performs well with very large meshes, outperforming other methods by generating continuous, high-fidelity forces free of discontinuities. It achieves high spatial accuracy by sampling a continuous implicit force function, enhancing the realism of haptic feedback in virtual environments. |
| 4. | Koukoulis E; Arvanitis G; Moustakas K Unleashing the Power of Generalized Iterative Closest Point for Swift and Effective Point Cloud Registration Conference 2024 IEEE International Conference on Image Processing (ICIP), IEEE 2024. @conference{koukoulis2024unleashing,
title = {Unleashing the Power of Generalized Iterative Closest Point for Swift and Effective Point Cloud Registration},
author = {Efthymios Koukoulis and Gerasimos Arvanitis and Konstantinos Moustakas},
url = {https://www.vvr.ece.upatras.gr/wp-content/uploads/sites/5/2024/11/Unleashing_the_Power_of_Generalized_Iterative_Closest_Point_for_Swift_and_Effective_Point_Cloud_Registration.pdf},
doi = {10.1109/ICIP51287.2024.10647551},
year = {2024},
date = {2024-09-27},
urldate = {2024-09-27},
booktitle = {2024 IEEE International Conference on Image Processing (ICIP)},
pages = {3403–3409},
organization = {IEEE},
abstract = {Point cloud registration is crucial for tasks like Simultaneous Localization and Mapping (SLAM) and 3D Reconstruction and relies on geometric information to achieve improved results. However, the complex cost functions involved in registration can pose challenges for general-purpose optimization algorithms. An efficient algorithm has been developed to solve the Generalized Least Squares problem within the Generalized Iterative Closest Point (GICP) framework, enabling faster and more accurate point cloud registration. The model is validated on a real-world autonomous driving dataset, where it outperforms existing methods in terms of both accuracy and real-time performance.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
Point cloud registration is crucial for tasks like Simultaneous Localization and Mapping (SLAM) and 3D Reconstruction and relies on geometric information to achieve improved results. However, the complex cost functions involved in registration can pose challenges for general-purpose optimization algorithms. An efficient algorithm has been developed to solve the Generalized Least Squares problem within the Generalized Iterative Closest Point (GICP) framework, enabling faster and more accurate point cloud registration. The model is validated on a real-world autonomous driving dataset, where it outperforms existing methods in terms of both accuracy and real-time performance. |
| 3. | Gkillas A; Anagnostopoulos C; Piperigkos N; Lalos A S A Real-time Explainable-by-design Super-Resolution Model for LiDAR SLAM Proceedings Article In: 2024 IEEE 7th International Conference on Industrial Cyber-Physical Systems (ICPS), pp. 1–8, IEEE 2024. @inproceedings{gkillas2024real,
title = {A Real-time Explainable-by-design Super-Resolution Model for LiDAR SLAM},
author = {Alexandros Gkillas and Christos Anagnostopoulos and Nikos Piperigkos and Aris S Lalos},
url = {https://www.vvr.ece.upatras.gr/wp-content/uploads/sites/5/2025/01/A_Real-time_Explainable-by-design_Super-Resolution_Model_for_LiDAR_SLAM.pdf, pdf/preprint},
year = {2024},
date = {2024-01-01},
urldate = {2024-01-01},
booktitle = {2024 IEEE 7th International Conference on Industrial Cyber-Physical Systems (ICPS)},
pages = {1–8},
organization = {IEEE},
abstract = {A novel optimization based deep learning architecture has been developed, utilizing the advantages of the range view domain. The mathematical connection between low and high-resolution images has been exploited by transforming 3D point clouds into 2D range images, towards a new optimization problem that combines learnable and handcrafted regularizers, to fully capture the inherent structure of the LiDAR data, i.e., the low-rank properties due to the strong correlations among neighboring points and the ring-like structure. The model leverages a deep learning architecture based on the Deep Unrolling strategy, offering both explainability and computational efficiency, and is validated on a real-world autonomous driving dataset, where it outperforms existing methods in terms of both accuracy and real-time performance. Operating at 50-100 frames per second (fps), the proposed solution significantly enhances real-time capabilities, a major improvement over current methods that fail to meet the critical requirement for real-time operation in SLAM applications.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
A novel optimization based deep learning architecture has been developed, utilizing the advantages of the range view domain. The mathematical connection between low and high-resolution images has been exploited by transforming 3D point clouds into 2D range images, towards a new optimization problem that combines learnable and handcrafted regularizers, to fully capture the inherent structure of the LiDAR data, i.e., the low-rank properties due to the strong correlations among neighboring points and the ring-like structure. The model leverages a deep learning architecture based on the Deep Unrolling strategy, offering both explainability and computational efficiency, and is validated on a real-world autonomous driving dataset, where it outperforms existing methods in terms of both accuracy and real-time performance. Operating at 50-100 frames per second (fps), the proposed solution significantly enhances real-time capabilities, a major improvement over current methods that fail to meet the critical requirement for real-time operation in SLAM applications. |
