Publication: Aplicación de ayuda para el reciclaje
Loading...
Official URL
Full text at PDC
Publication Date
2021-06
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
Con el incremento de los envases de un solo uso y el aumento de los tipos de materiales, se ha observado un crecimiento masivo en los residuos, y estos, al no ser gestionados adecuadamente, han provocado un gran impacto medioambiental. Para hacerle frente a este problema una acción sencilla pero efectiva es separar los residuos en el origen. No obstante, con la diversidad de materiales existentes actualmente puede resultar complicado saber cómo desecharlos todos correctamente. Este Trabajo de Fin de Grado busca ofrecer, mediante una aplicación móvil, una pequeña ayuda para solventar estas dudas a personas que lo puedan necesitar.
Para ello, se ha querido aprovechar el gran avance que se ha vivido en visión computacional durante los últimos años y desarrollar una aplicación de identicación de objetos. Para esto se ha utilizado TensorFlow Lite, cuyas librerías facilitan tanto el entrenamiento de la red neuronal necesaria y la generación del modelo, como su importación en aplicaciones móviles. Asimismo, para facilitar la tarea de obtención de datos para el entrenamiento, se ha desarrollado una aplicación de generación de imágenes sintéticas, las cuales se generan a partir de modelos tridimensionales y son utilizadas posteriormente para el entrenamiento de la red.
Todo esto en conjunto permite obtener como resultado una aplicación móvil de identicación de materiales que ha sido entrenada utilizando imágenes generadas sintéticamente.
With the increase of single-use packaging and the different types of materials, there has been a massive growth in waste. Inappropiate management of these wastes have caused a huge environmental impact. To deal with this problem, one simple but effective action is to separate the waste at the source point. However, with the diversity of materials available today it can be difficult to know how to correctly dispose all of them. This final degree project seeks to offer help to solve the doubts through a mobile application. To do so, we wanted to take advantage of the great progress that has been made in computer vision in recent years and develop an application that identifies objects. To achieve this, TensorFlow Lite has been used, whose libraries provides an easy way for training the needed neural network and generate its model, as well as its import into mobile applications. Likewise, to ease the task of obtaining the necessary datasets for training, an application that generates synthetic images from three-dimensional models has been developed. These images are subsequently used for the neuronal network training. As a result we obtain a mobile application which identifies different materials trained using synthetic images.
With the increase of single-use packaging and the different types of materials, there has been a massive growth in waste. Inappropiate management of these wastes have caused a huge environmental impact. To deal with this problem, one simple but effective action is to separate the waste at the source point. However, with the diversity of materials available today it can be difficult to know how to correctly dispose all of them. This final degree project seeks to offer help to solve the doubts through a mobile application. To do so, we wanted to take advantage of the great progress that has been made in computer vision in recent years and develop an application that identifies objects. To achieve this, TensorFlow Lite has been used, whose libraries provides an easy way for training the needed neural network and generate its model, as well as its import into mobile applications. Likewise, to ease the task of obtaining the necessary datasets for training, an application that generates synthetic images from three-dimensional models has been developed. These images are subsequently used for the neuronal network training. As a result we obtain a mobile application which identifies different materials trained using synthetic images.
Description
Trabajo de Fin de Grado en Desarrollo de Videojuegos, Facultad de Informática UCM, Departamento de Ingeniería del Software e Inteligencia Artificial, Curso 2020/2021