A5090137141- 3D Shape Modeling and Analysis
- Image Processing and 3D Reconstruction
- Computer Graphics and Visualization Techniques
- 3D Surveying and Cultural Heritage
- Human Pose and Action Recognition
- Hand Gesture Recognition Systems
- Infrared Thermography in Medicine
- Infrared Target Detection Methodologies
- Gait Recognition and Analysis
- Image Retrieval and Classification Techniques
- Image and Object Detection Techniques
- Hydrology and Sediment Transport Processes
- Thermography and Photoacoustic Techniques
- Landslides and related hazards
- Remote Sensing and LiDAR Applications
University of Patras
2020-2024
Vrije Universiteit Brussel
2023-2024
In this paper we delve into the properties of transformers, attained through self-supervision, in point cloud domain. Specifically, evaluate effectiveness Masked Autoencoding as a pretraining scheme, and explore Momentum Contrast an alternative. our study investigate impact data quantity on learned features, uncover similarities transformer's behavior across domains. Through comprehensive visualizations, observe that transformer learns to attend semantically meaningful regions, indicating...
Gesture recognition is a tool to enable novel interactions with different techniques and applications, like Mixed Reality Virtual environments. With all the recent advancements in gesture from skeletal data, it still unclear how well state-of-the-art perform scenario using precise motions two hands. This paper presents results of SHREC 2024 contest organized evaluate methods for their highly similar hand spatial coordinate data both The task 7 motion classes given coordinates frame-by-frame...
We propose a novel point cloud U-Net diffusion architecture for 3D generative modeling capable of generating high-quality and diverse shapes while maintaining fast generation times. Our network employs dual-branch architecture, combining the high-resolution representations points with computational efficiency sparse voxels. fastest variant outperforms all non-diffusion approaches on unconditional shape generation, most popular benchmark evaluating models, our largest model achieves...
In this paper we study the problem of shape part retrieval in point cloud domain. Shape methods literature rely on presence an existing query object, but what if are looking for is not available? We present Part Retrieval Pipeline (PReP), a pipeline that creatively utilizes metric learning techniques along with trained classification model to measure suitability potential replacement parts from database, as application scenario targeting circular economy. Through innovative training...
In this paper we delve into the properties of transformers, attained through self-supervision, in point cloud domain. Specifically, evaluate effectiveness Masked Autoencoding as a pretraining scheme, and explore Momentum Contrast an alternative. our study investigate impact data quantity on learned features, uncover similarities transformer's behavior across domains. Through comprehensive visualiations, observe that transformer learns to attend semantically meaningful regions, indicating...