A5031450889- Visual Attention and Saliency Detection
- Advanced Image and Video Retrieval Techniques
- Video Surveillance and Tracking Methods
- Water Quality Monitoring Technologies
- Domain Adaptation and Few-Shot Learning
- EEG and Brain-Computer Interfaces
- Multimodal Machine Learning Applications
- Neural dynamics and brain function
- Identification and Quantification in Food
- Advanced Neural Network Applications
- Privacy-Preserving Technologies in Data
- COVID-19 diagnosis using AI
- AI in cancer detection
- Generative Adversarial Networks and Image Synthesis
- Advanced Vision and Imaging
- Human Pose and Action Recognition
- Computer Graphics and Visualization Techniques
- Image and Video Quality Assessment
- Radiomics and Machine Learning in Medical Imaging
- Image Enhancement Techniques
- Robotics and Sensor-Based Localization
- Face Recognition and Perception
- Speech and Audio Processing
- Robotic Path Planning Algorithms
- Artificial Intelligence in Healthcare and Education
University of Catania
2016-2025
Gesellschaft für Anlagen und Reaktorsicherheit
2013-2023
Royal Military Academy
2020
Medica (Italy)
2017
Azienda Ospedaliera di Cosenza
2016
University of Edinburgh
2012
xWhat if we could effectively read the mind and transfer human visual capabilities to computer vision methods? In this paper, aim at addressing question by developing first object classifier driven brain signals. particular, employ EEG data evoked stimuli combined with Recurrent Neural Networks (RNN) learn a discriminative activity manifold of categories in reading effort. Afterward, learned machines training Convolutional Network (CNN)-based regressor project images onto manifold, thus...
We present a research tool that supports marine ecologists' by allowing analysis of long-term and continuous fish monitoring video content. The can be used for instance to discover ecological phenomena such as changes in abundance species composition over time area. Two characteristics set our system apart from traditional data collecting processing methods. First, the recording results enormous volumes data. Currently around year recordings (containing 4 million observations) have been...
Recent advancements in generative adversarial networks (GANs), using deep convolutional models, have supported the development of image generation techniques able to reach satisfactory levels realism. Further improvements been proposed condition GANs generate images matching a specific object category or short text description. In this work, we build on latter class approaches and investigate possibility driving conditioning process by means brain signals recorded, through an...
This work presents a novel method of exploring human brain-visual representations, with view towards replicating these processes in machines. The core idea is to learn plausible computational and biological representations by correlating neural activity natural images. Thus, we first propose model, EEG-ChannelNet, brain manifold for EEG classification. After verifying that visual information can be extracted from data, introduce multimodal approach uses deep image encoders, trained siamese...
Reading the human mind has been a hot topic in last decades, and recent research neuroscience found evidence on possibility of decoding, from neuroimaging data, how brain works. At same time, rediscovery deep learning combined to large interest scientific community generative methods enabled generation realistic images by data distribution noise. The quality generated increases when input conveys information visual content images. Leveraging these trends, this paper we present an approach...
In this paper we present an approach for segmenting objects in videos taken complex scenes with multiple and different targets. The method does not make any specific assumptions about the relies on how are perceived by humans according to Gestalt laws. Initially, rapidly generate a coarse foreground segmentation, which provides predictions motion regions analyzing superpixel segmentation changes consecutive frames. We then exploit these location priors refine initial optimizing energy...
Abstract In this work, we propose a 3D fully convolutional architecture for video saliency prediction that employs hierarchical supervision on intermediate maps (referred to as conspicuity ) generated using features extracted at different abstraction levels. We provide the base learning mechanism with two techniques domain adaptation and domain-specific . For former, encourage model unsupervisedly learn general gradient reversal multiple scales, enhance generalization capabilities datasets...
Volcano-seismic signals can help for volcanic hazard estimation and eruption forecasting. However, the underlying mechanism their low frequency components is still a matter of debate. Here, we show signatures dynamic strain records from Distributed Acoustic Sensing in frequencies at Vulcano Island, Italy. Signs unrest have been observed since September 2021, with CO2 degassing occurrence long period very events. We interrogated fiber-optic telecommunication cable on-shore off-shore linking...
The integration of Artificial Intelligence, particularly foundation models and modern Transformer-based architectures, opens up new frontiers for seismic monitoring. In this work, we propose a comprehensive AI-driven framework detection phase picking events. These are designed to exploit the capabilities advanced AI techniques tackle challenges posed by high-frequency, high-density data, noisy environments typically associated with monitoring technologies like Distributed Acoustic Sensing...
In this paper we present SurfaceNet, an approach for estimating spatially-varying bidirectional reflectance distribution function (SVBRDF) material properties from a single image. We pose the problem as image translation task and propose novel patch-based generative adversarial network (GAN) that is able to produce high-quality, high-resolution surface maps. The employment of GAN paradigm has twofold objective: 1) allowing model recover finer details than standard models; 2) reducing domain...