Now showing items 1-13 of 13

    • Aliasing is a driver of adversarial attacks 

      Rodríguez Muñoz, Adrián (Universitat Politècnica de Catalunya, 2022-05)
      Bachelor thesis
      Restricted access - confidentiality agreement
    • Aligning books and movies through cross-modal neural networks 

      Hernandez Caralt, Mireia (Universitat Politècnica de Catalunya, 2021-05-27)
      Bachelor thesis
      Open Access
      Covenantee:   Massachusetts Institute of Technology
      Aquesta tesi explora les millors pràctiques per alinear llibres amb la seva adaptació cinematogràfica utilitzant xarxes neuronals cros-modals. Primer col·lectem dades per la classificació de paràgrafs basada en el seu ...
    • Detección de objetos y anotación de imágenes en el iPhone 

      Blanco Almazan, M. Dolores (Universitat Politècnica de Catalunya, 2013-02-25)
      Master thesis (pre-Bologna period)
      Open Access
      Covenantee:   Massachusetts Institute of Technology
      [ANGLÈS] This project is about object detection on a mobile device like iPhone. The iPhone tool allows users to take pictures, annotate them with bounding boxes and object tags and uploading the data to the server. With ...
    • DetectMe: object detection on the iPhone 

      Mingot Hidalgo, Josep Marc (Universitat Politècnica de Catalunya, 2013-12-19)
      Master thesis (pre-Bologna period)
      Open Access
      Covenantee:   Massachusetts Institute of Technology
      [ANGLÈS] DetectMe is one of the firsts apps to do the complete process of object detection on the iOS devices. With the app, the user is able to train a detector to detect any kind of objects and later on execute it on ...
    • Frames in places: visual common sense knowledge in context 

      Puig Fernández, Xavier (Universitat Politècnica de Catalunya, 2016-05-26)
      Bachelor thesis
      Open Access
      Covenantee:   Massachusetts Institute of Technology
      One of the goals of Computer Vision is to be able to understand real-world situations at the level humans can do. In order to adapt to a constantly changing world, we use common sense to comprehend and react to events even ...
    • How concepts emerge in neural networks 

      Surís Coll-Vinent, Dídac (Universitat Politècnica de Catalunya, 2018-10-17)
      Master thesis
      Restricted access - author's decision
      Covenantee:   Massachusetts Institute of Technology
      Deep learning models, and more specifically computer vision systems, have achieved great results in recent years. However, the interpretability and understanding of these models is still in its early stages. Interpretability ...
    • Image-based recognition of natural disasters and other events that might require human assistance 

      Marzo Grimalt, Núria (Universitat Politècnica de Catalunya, 2019-05-28)
      Bachelor thesis
      Open Access
      Covenantee:   Massachusetts Institute of Technology
      The work is going to consist on building a dataset with images from incidents such as fires, earthquakes, floodings, contamination… in different scenarios. The data needs to be cleaned and is going to be annotated using ...
    • Learning implicit shape representations from tactile sensing 

      Ventura Ripol, Lucas (Universitat Politècnica de Catalunya, 2021-07)
      Master thesis
      Restricted access - author's decision
      Covenantee:   Massachusetts Institute of Technology
    • Light field inference from diffuse shadows 

      Baradad Jurjo, Manel (Universitat Politècnica de Catalunya, 2017-10)
      Master thesis
      Restricted access - author's decision
      In this thesis we propose a method to infer light fields from shadows casted over diffuse reflectors. Despite shadows can be seen as projections of a 4D function (i.e.\ the light field) into a lower dimensional space (i.e.a ...
    • Object detection and recognition: from saliency prediction to one-shot trained detectors 

      Recasens Continente, Adrià (Universitat Politècnica de Catalunya, 2014-09-03)
      Master thesis (pre-Bologna period)
      Open Access
      Covenantee:   Massachusetts Institute of Technology
      [ANGLÈS] Computer vision capabilities have started to become available in smart devices this last years. The rapid growth of the smartphone world along with the big advance of the computer vision field in the last years ...
    • Semantic Label Sharing for Semi-Supervised learning with large datasets 

      Jiménez Bernal, Héctor (Universitat Politècnica de Catalunya / Massachusetts Institute of Technology, 2011-12)
      Master thesis (pre-Bologna period)
      Open Access
      Covenantee:   Massachusetts Institute of Technology
      [ANGLÈS] In an object recognition scenario with tens of thousands of categories, even a small number of labels per category leads to a very large number of total labels required. The proposed methodology consists on label ...
    • Understanding human interaction in TV shows 

      Felip Díaz, Bernat (Universitat Politècnica de Catalunya, 2020-10-16)
      Bachelor thesis
      Open Access
      Covenantee:   Massachusetts Institute of Technology
      En aquest projecte, diversos models d’aprenentatge profund d’última generació s'han fet servir per extreure informació útil de programes de televisió per en un futur poder analitzar les interaccions de les persones que es ...
    • Vision and language: from visual question answering to domain-invariant representations 

      Castrejón Subira, Lluis Enric (Universitat Politècnica de Catalunya, 2015-05-18)
      Master thesis (pre-Bologna period)
      Restricted access - author's decision
      Covenantee:   Massachusetts Institute of Technology
      [ANGLÈS] The field of computer vision has radically evolved in the last few years due to the success of deep artificial neural networks. Current models have achieved remarkable performance in hallmark vision tasks such as ...