Search

Esta memoria describe cómo se ha desarrollado una aplicación de visión por computador basada en la placa de desarrollo Sony Spresense. Toma como punto de referencia el estudio previo realizado por el estudiante Yeray Navarro en su Trabajo de Fin de Grado, Placa programable Spresense de Sony. Desarroyo de una aplicación de visión por computador" y los modelos de visión por computador definidos en él. A lo largo del trabajo se definen diversas etapas que engloban el desarrollo de la aplicación, desde las etapas iniciales donde se concretan los objetivos, hasta el uso práctico de la aplicación. El desarrollo de este trabajo se pretende aplicar a la mejora de la prótesis de brazo diseñada y fabricada por el equipo de ARM2u, proyecto universitario impulsado por la Escuela Técnica Superior de Ingeniería Industrial de Barcelona, ETSEIB, de la Universidad Politécnica de Cataluña, UPC. El proceso de creación de la aplicación se realiza de forma segmentada, empezando por un código base embebido en la placa de desarrollo de captura y clasificación de imágenes, pasando por una primera versión de la aplicación unida al código de movimiento de la prótesis y finalizando con una optimización de la misma. Al concluir este trabajo se presenta una aplicación embebida en la placa de desarrollo Sony Spresense, cuyas funcionalidades son procesadas completamente por el procesador incorporado en la placa. Los resultados de diferentes pruebas realizadas sobre la aplicación permiten demostrar su funcionalidad y validar el modelo de aplicación creado. Se configura como una aplicación que permite el movimiento natural de la prótesis pero, a voluntad del usuario, permite el cierre de la pinza al tamaño del objeto resultante de la clasificación de imágenes Aquesta memòria descriu com s' ha desenvolupat una aplicació de visió per computador basada en la placa de desenvolupament Sony Spresense. Pren com a punt de referència l'estudi previ realitzat per l'estudiant Yeray Navarro en el seu Treball de Fi de Grau, "Placa Programable Spresense de Sony. Desarrollo de una aplicación de visión por computador" i els models de visió per computador definits en ell. Al llarg del treball es defineixen diverses etapes que engloben el desenvolupament de l'aplicació, des de les etapes inicials on es fixen els objectius, fins a l'ús pràctic de l'aplicació. El desenvolupament d' aquest treball va dirigit a la millora de la pròtesi de braç dissenyada i fabricada per l'equip ARM2u, projecte universitari impulsat per l'Escola Tècnica Superior d'Enginyeria Industrial de Barcelona, de la Universitat Politècnica de Catalunya, UPC. El procés de creació de l'aplicació es realitza de forma segmentada, començant per un codi base embegut a la placa de desenvolupament de captura i classificació d'imatges, passant per una primera versió de l'aplicació unida al codi de moviment de la pròtesi i finalitzant amb una optimització de la mateixa. En concloure aquest treball es presenta una aplicació embeguda a la placa de desenvolupament Sony Spresense, les funcionalitats de la qual són processades completament pel processador incorporat a la placa. Els resultats de diferents proves realitzades sobre l'aplicació permeten demostrar la seva funcionalitat i validar el model d' aplicació creat. Es configura com una aplicació que permet el moviment natural de la pròtesi però que, a voluntat de l'usuari, permet el tancament de la pinça a la mida de l' objecte resultant de la classificació d'imatges This report describes how a computer vision application based on the Sony Spresense development board has been developed. It takes as a point of reference the previous study carried out by the student Yeray Navarro in his Final Degree Project, "Placa Programable Spresense de Sony. Desarrollo de una aplicación de visión por computador" and the computer vision models defined in it. Throughout the work, various stages are defined that encompass the development of the application, from the initial stages where the objectives to be met are set to the practical use of the application. The development of this work is aimed at improving the arm prosthesis designed and manufactured by the ARM2u team, pushed forward by the Escola Tècnica Superior d'Enginyeria Industrial de Barcelona, ETSEIB, of the Universitat Politècnica de Catalunya, UPC. The process of creating the application is carried out in a segmented way, starting with a base code embedded in the image capture and classification development board, passing through a first version of the application together with the movement code of the prosthesis and ending with an optimization of it. At the conclusion of this work, an application embedded in the Sony Spresense development board is presented, whose functionalities are completely processed by the processor incorporated in the board. The results of different tests carried out on the application allow to demonstrate its functionality and validate the application model created. It is configured as an application that allows the natural movement of the prosthesis that, at the user's will, allows the closure of the clamp to the size of the object resulting from the classification of image

Pixel-wise image segmentation is key for many Computer Vision applications. The training of deep neural networks for this task has expensive pixel-level annotation requirements, thus, motivating a growing interest on synthetic data to provide unlimited data and its annotations. In this paper, we focus on the generation and application of synthetic data as representative training corpuses for semantic segmentation of urban scenes. First, we propose a synthetic data generation protocol, which identifies key features affecting performance and provides datasets with variable complexity. Second, we adapt two popular weakly supervised domain adaptation approaches (combined training, fine-tuning) to employ synthetic and real data. Moreover, we analyze several backbone models, real/synthetic datasets and their proportions when combined. Third, we propose a new curriculum learning strategy to employ several synthetic and real datasets. Our major findings suggest the high performance impact of pace and order of synthetic and real data presentation, achieving state of the art results for well-known models. The results by training with the proposed dataset outperform popular alternatives, thus demonstrating the effectiveness of the proposed protocol. Our code and dataset are available at http://www-vpu.eps.uam.es/publications/WSDA_semantic/, Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature. This work is part of the preliminary tasks related to the SEGA-CV (TED2021-131643A-I00) and the HVD (PID2021-125051OB-I00) projects funded by the Ministerio de Ciencia e Innovacion of the Spanish Government

In this work we explore enhancing performance of reinforcement learning algorithms in video game environments by feeding it better, more relevant data. For this purpose, we use semantic segmentation to transform the images that would be used as input for the reinforcement learning algorithm from their original domain to a simplified semantic domain with just silhouettes and class labels instead of textures and colors, and then we train the reinforcement learning algorithm with these simplified images. We have conducted different experiments to study multiple aspects: feasibility of our proposal, and potential benefits to model generalization and transfer learning. Experiments have been performed with the Super Mario Bros video game as the testing environment. Our results show multiple advantages for this method. First, it proves that using semantic segmentation enables reaching higher performance than the baseline reinforcement learning algorithm without modifying the actual algorithm, and in fewer episodes; second, it shows noticeable performance improvements when training on multiple levels at the same time; and finally, it allows to apply transfer learning for models trained on visually different environments. We conclude that using semantic segmentation can certainly help reinforcement learning algorithms that work with visual data, by refining it. Our results also suggest that other computer vision techniques may also be beneficial for data prepossessing. Models and code will be available on github upon acceptance, Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature. This work is part of the preliminary tasks related to the Harvesting Visual Data (HVD) project (PID2021- 125051OB-I00) funded by the Ministerio de Ciencia e Innovacin of the Spanish Government

This paper describes the scientific achievements of a collaboration between a research group and the waste management division of a company. While these results might be the basis for several practical or commercial developments, we here focus on a novel scientific contribution: a methodology to automatically generate geo-located waste container maps. It is based on the use of Computer Vision algorithms to detect waste containers and identify their geographic location and dimensions. Algorithms analyze a video sequence and provide an automatic discrimination between images with and without containers. More precisely, two state-of-the-art object detectors based on deep learning techniques have been selected for testing, according to their performance and to their adaptability to an on-board real-time environment: EfficientDet and YOLOv5. Experimental results indicate that the proposed visual model for waste container detection is able to effectively operate with consistent performance disregarding the container type (organic waste, plastic, glass and paper recycling,…) and the city layout, which has been assessed by evaluating it on eleven different Spanish cities that vary in terms of size, climate, urban layout and containers’ appearance, This work has been supported by URBASER S.A. and the Universidad Autonoma ´ de Madrid under project REVGA of the ”Segunda Edicion ´ del Programa de Fomento de la Transferencia del Conocimiento” call

En este trabajo de Fin de Grado se expone un proyecto de innovación relacionado con el aprendizaje de la lengua inglesa a través de la gamificación en el aula de Educación Primaria. Esta propuesta se centra principalmente en el aprendizaje de segundas lenguas a través de metodologías vanguardistas adaptadas al curso de 3º de primaria y a las necesidades de aprendizaje de cada alumno. Así, la gamificación se manifiesta como un procedimiento progresivo que facilita la adquisición de determinados contenidos educativos. Por otro lado, la enseñanza de la lengua inglesa es fundamental hoy en día en este mundo globalizado, por lo que es necesario adquirir estos conocimientos desde edades tempranas. Durante el desarrollo de este trabajo definimos la gamificación y cómo aplicarla adecuadamente para conseguir resultados en un aula de primaria. Además, se presenta una propuesta de innovación a través del desarrollo de un escape room, realizado con herramientas TIC y recursos analógicos. This Final Degree project sets out an innovation project related to English language learning through gamification in the Primary Education classroom. This proposal focuses mainly on the learning of second languages through avant-garde methodologies, adapted to the 3rd year of primary school and the learning needs of each student. Thus, gamification manifests itself as a progressive procedure that facilitates the acquisition of certain educational content. On the other hand, the teaching of the English language is essential nowadays in this globalised world, which is why it is necessary to acquire this knowledge from an early age. During the development of this project, we define gamification and how to apply it appropriately to achieve results in a primary school classroom. In addition, an innovation proposal is presented through the development of an escape room, carried out with ICT tools and analogue resources.

Trabajo de fin de Grado. Grado en Maestro de Educación Primaria (E.U. de Magisterio Zamora). Curso académico 2018, [ES] Este trabajo confirma la dificultad para el uso de herramientas en el aula en la enseñanza del alemán , y además, adolece de manuales cuyas unidades didácticas incidan en la competencia comunicativa y en el aspecto sociocultural del país,para ello, se presenta una propuesta metodológica mediante la elaboración de materiales de trabajo,, láminas y actividades complementarias a los libros de texto., [EN] This work confirms the difficulty for the use of tools in the classroom in the teaching of German, and also suffers from manuals whose teaching units affect the communicative competence and the socio-cultural aspect of the country, for this, a methodological proposal is presented through the elaboration of work materials, sheets and complementary activities to textbooks.

Tesis doctoral inédita leída en la Universidad Autónoma de Madrid. Escuela Politécnica Superior, Departamento de Tecnología Electrónica y de las Comunicaciones. Fecha de lectura: junio de 2013

Master in Deep Learning for Audio and Video Signal Processing, This Master Thesis is contextualized in the context of the need to count containersin different cities in Spain for garbage collection. In order to perform this countingautomatically, to save time and resources, it is proposed to use the technique of objectdetection on images captured from a car.At the research proposal level, the hypothesis is as follows: ”it is possible to createa model/system that takes advantage of the knowledge acquired from the cubes of acity to facilitate learning in new cities, without significantly worsening its results inthe previously known city”. To achieve this, one has to deal with a problem deriveddirectly from the training algorithm of a neural network, the Catastrophic Forgetting.This problem implies that, once a model with certain weights is trained to solve a newtask (e.g., recognition of a new class) its results will be worse on the previously knowntask/class.In order to try to solve this problem, a current called Continual Learning (alsoknown as Incremental Learning or Lifelong Learning) arises within the area of Arti-ficial Intelligence, which introduces techniques based on reproduction, regularizationor isolation of parameters, among others, to try to mitigate the effect of CatastrophicForgetting in the results of the model, retaining the knowledge prior to the retrainingof the model with the new data.Specifically, in this project we have worked to achieve a system that is capable ofdetecting containers of new cities without losing accuracy in the detection of previouslyknown cities. To this end, we have carried out annotation tasks, adaptation of a GitHubrepository where the EfficientDet model is used for object detection, measurement andmitigation of catastrophic forgetting and finally creation of an interface to facilitatethe use of the application to a potential user.

This Master Thesis focuses on applying semantic segmentation, a computer vision technique, with the objective of improving the performance of deep-learning reinforcement models, and in particular, the performance over the original Super Mario Bros videogame. While humans can play a stage from a videogame like Super Mario Bros, and quickly identify from the elements in the screen what object is the character they are playing with, what are enemies and what elements are obstacles, this is not the case for neural networks, as they require a certain training to understand what is displayed in the screen. Using semantic segmentation, we can heavily simplify the frames from the videogame, and reduce visual information of elements in the screen to class and location, which is the most relevant information required to complete the game. In this work, a synthetic dataset generator that simulates frames from the Super Mario Bros videogame has been developed. This dataset has been used to train semantic segmentation deep-learning models which have been incorporated to a deep reinforcement learning algorithm with the objective of improving the performance of it. We have found that applying semantic segmentation as a frame processing method can actually help reinforcement learning models to train more efficiently and with better generalization. These results also suggest that there could be other computer vision techniques, like object detection or tracking, that could be found useful to help with the training of reinforcement learning algorithms, and they could be an interesting topic for future research.