A5005762795- Advanced Image and Video Retrieval Techniques
- Robotics and Sensor-Based Localization
- Advanced Vision and Imaging
- Advanced Neural Network Applications
- Video Surveillance and Tracking Methods
- Image and Object Detection Techniques
- Human Pose and Action Recognition
- Anomaly Detection Techniques and Applications
- Video Analysis and Summarization
- Time Series Analysis and Forecasting
- Image Retrieval and Classification Techniques
- Sports Analytics and Performance
- Human Motion and Animation
- Optical measurement and interference techniques
- Image Processing and 3D Reconstruction
- 3D Shape Modeling and Analysis
- Machine Learning and Algorithms
- Sports Dynamics and Biomechanics
- Machine Learning and Data Classification
- Digital Media Forensic Detection
- 3D Surveying and Cultural Heritage
- Image Processing Techniques and Applications
- Spectroscopy Techniques in Biomedical and Chemical Research
- Autonomous Vehicle Technology and Safety
- Infrared Target Detection Methodologies
Universitat Politècnica de Catalunya
2015
Institut de Robòtica i Informàtica Industrial
2015
Universitat Pompeu Fabra
2011-2014
FC Barcelona
2010
Universitat Autònoma de Barcelona
2005-2009
Laboratoire d'Informatique de Paris-Nord
2006
Deep learning has revolutionalized image-level tasks such as classification, but patch-level tasks, correspondence, still rely on hand-crafted features, e.g. SIFT. In this paper we use Convolutional Neural Networks (CNNs) to learn discriminant patch representations and in particular train a Siamese network with pairs of (non-)corresponding patches. We deal the large number potential combination stochastic sampling training set an aggressive mining strategy biased towards patches that are...
We propose a real-time, robust to outliers and accurate solution the Perspective-n-Point (PnP) problem. The main advantages of our are twofold: first, it in- tegrates outlier rejection within pose estimation pipeline with negligible computational overhead, sec- ond, its scalability arbitrarily large number correspon- dences. Given set 3D-to-2D matches, we formulate problem as low-rank homogeneous sys- tem where lies on 1D null space. Outlier correspondences those rows linear system which...
In this paper we propose a novel framework for learning local image descriptors in discriminative manner. For purpose explore siamese architecture of Deep Convolutional Neural Networks (CNN), with Hinge embedding loss on the L2 distance between descriptors. Since uses pairs rather than single patches to train, there exist large number positive samples and an exponential negative samples. We space stochastic sampling training set, combination aggressive mining strategy over both which denote...
Trabajo presentado a la 25th British Machine Vision Conference (BMVC), celebrada en Nottingham (UK) del 1 al 5 de septiembre 2014.-- Este item (excepto textos e imagenes no creados por el autor) esta sujeto una licencia Creative Commons: Attribution-NonCommercial-NoDerivs 3.0 Spain.
Millimetric Waves Images (MMW) are becoming more and useful in the passive detection of threaten objects based on plastic substances as explosives or sharp/cutting weapons. Our goal is to achieve segmentation body concealed threats dealing with inherent problems this type images: noise, low resolution intensity inhomogeneity. In work we present results applying Iterative Steering Kernel Regression (ISKR) method for denoising Local Binary Fitting (LBF) order correctly segment bodies over a...
We propose a robust and efficient method to estimate the pose of camera with respect complex 3D textured models environment that can potentially contain more than 100; 000 points. To tackle this problem we follow top down approach where combine high-level deep network classifiers low level geometric approaches come up solution is fast, accurate. Given an input image, initially use pre-trained compute rough estimation pose. This initial constrains number model points be seen from viewpoint....
In tracking tasks, representing a target region as weighted histogram has opened possibilities which led to excellent results, mean shift or camshift algorithms. This representation is extracted from the image by giving weights with kernels and it depends on properties of kernels. By first order Taylor approximation histograms possible perform using several kernels, interpreted different sources information. improves gives more flexibility when facing problems tracking, occlusions, model...
Motion prediction in soccer involves capturing complex dynamics from player and ball interactions. We present FootBots, an encoder-decoder transformer-based architecture addressing motion conditioned through equivariance properties. FootBots captures temporal social using set attention blocks multi-attention block decoder. Our evaluation utilizes two datasets: a real dataset tailored synthetic one. Insights the highlight effectiveness of FootBots' mechanism significance prediction. Empirical...
Understanding trajectories in multi-agent scenarios requires addressing various tasks, including predicting future movements, imputing missing observations, inferring the status of unseen agents, and classifying different global states. Traditional data-driven approaches often handle these tasks separately with specialized models. We introduce TranSPORTmer, a unified transformer-based framework capable all showcasing its application to intricate dynamics sports like soccer basketball. Using...
In this paper we present a novel scale invariant interest point detector of blobs which incorporates the idea blob movement along scales. This trajectory through space is shown to be valuable information in order estimate most stable locations and scales points. Our evaluates points terms their self its evolution avoiding redundant detections. Moreover, differential geometry view understand how can detected. We propose analyze gaussian curvature classify image regions as elliptical (blobs)...
This paper addresses the problem of tracking a target in an IR video sequence using kernel based histogram representation target. In this field, gradient ascent methods have demonstrated useful results with weighted kernels and particular Mean Shift is currently most commonly used scale method. Our approximation follows work made by Hager, that uses SSD objective function (derived from Matusita metric) combines it Newton-like maximization method, resulting fast system. An important property...