H2020HuPBA participation in international projects (European Comission, selection):

Hermes 2007-2010 // WIDER – Green Growing of SMEs: Innovation and Development in the Energy Sector in the Mediterranean Area 2015 // TeSLA 2016-2018 // SME Disriptive OnLine Stylist Engine by Value Agents 2015 // SME Automatic Digital Biometry Analysis System for musculoskeletal disorders rehabilitation (ADIBAS) 2017 // SEE.4C – SpatiotEmporal ForEcasting: Coopetition to meet Current Cross-modal Challenges 2016-2017 // Cost Action: Integrating Vision & Language 2014-2018 // iCARE 2014-2018 // COMPASS No 710543 2017-2020 // Innolabs No 691556 2017-2019


Multi-class Visual Object and Pattern Classification via Error Correcting Output Codes

ECOCLib: Open Source Error-Correcting Output Codes library


Human Pose Recovery and Behavior Analysis

Web of the project


This smart phone application is able to detect text in complex scenes, remove text from images, reconstruct background, apply OCR, translate text, and draw the new text as part of the image.

Execution example video


Robot navigation and interaction

Case Base Reasoning: In this project we combine feature weighting and Reinforcement Learning approaches to guide the user to a faster convergence of product location in a case base data set.


Medical Image Segmentation
Automatic segmentation of artifacts in X-Ray, CT, and MRI images using Graph Cut Theory.


Mobile Mapping system: the position and orientation of the different traffic signs are measured with video cameras fixed on a moving vehicle. Signs are detected by means of a cascade of classifiers using Adaboost. Multi-class traffic sign classification is then performed using of Error Correcting Output Codes. Collaborators: Intitut Cartogràfic de Catalunya 



This system deals with the problem of multi-class airplain recognition within airport environments from video sequences, focusing on the segmentation, feature extraction, and learning stages. Collaborators: AENA


Volum visualization: classification and segmentation of large medical data sets – GPGPU Multi-class volume classification video


Contextual learning: Multi-class Multi-scale Stacked Sequential Learning


 In this project a mobile system takes pictures of urban street floor. The purpose of the project is first to detect the presence of residues on the floor, the origin of those residues (organic/inorganic) and the categorization of each individual residue into more explicit categories (e.g. paper, cigarette, can, etc).
Collaborators: INYPSA