A5088282028- Human Pose and Action Recognition
- Video Surveillance and Tracking Methods
- Advanced Vision and Imaging
- Image Enhancement Techniques
- Advanced Image Processing Techniques
- Multimodal Machine Learning Applications
- Hand Gesture Recognition Systems
- EEG and Brain-Computer Interfaces
- E-Learning and Knowledge Management
- Biomedical and Engineering Education
- Blind Source Separation Techniques
- Human Motion and Animation
- Image Retrieval and Classification Techniques
- Anomaly Detection Techniques and Applications
- Robotics and Sensor-Based Localization
- Optical measurement and interference techniques
- Video Analysis and Summarization
- 3D Surveying and Cultural Heritage
- Privacy, Security, and Data Protection
- Neural Networks and Applications
- Advanced Image and Video Retrieval Techniques
- 3D Printing in Biomedical Research
- Ethics and Social Impacts of AI
University of Applied Sciences Utrecht
2024
Utrecht University
2013-2018
The introduction of deep learning and transfer techniques in fields such as computer vision allowed a leap forward the accuracy image classification tasks. Currently there is only limited use neuroscience. challenge using methods to successfully train models neuroscience, lies complexity information that processed, availability data cost producing sufficient high quality annotations. Inspired by its application vision, we introduce on electrophysiological enable training model with amounts...
We introduce a novel spatiotemporal deformable part model for the localization of fine-grained human interactions two persons in unsegmented videos. Our approach is first to classify and additionally provide temporal spatial extent interaction video. To this end, our models contain detectors that support different scales as well types feature descriptors, which are combined single graph. This allows us detailed coordination between people terms body pose motion. demonstrate helps avoid...
The intersection of physics-based vision and deep learning presents an exciting frontier for advancing computer technologies. By leveraging the principles physics to inform enhance models, we can develop more robust accurate systems. Physics-based aims invert processes recover scene properties such as shape, reflectance, light distribution, medium from images. In recent years, has shown promising improvements various tasks, when combined with vision, these approaches robustness accuracy This...
We model dyadic (two-person) interactions by discriminatively training a spatio-temporal deformable part of fine-grained human interactions. All involve at most two persons. Our models are capable localizing in unsegmented videos, marking the interest space and time. contributions as follows: First, we create that localizes Second, our use multiple pose motion features per part. Third, experiment with different ways discriminatively. When testing on target class achieve mean average...
In social settings, people interact in close proximity. When analyzing such encounters from video, we are typically interested distinguishing between a large number of different interactions. Here, address training deformable part models (DPMs) for the detection interactions both space and time. consider interaction classes, face two challenges. First, need to distinguish that visually more similar. Second, it becomes difficult obtain sufficient specific examples each class. this paper,...