Use this link to cite:
http://hdl.handle.net/2183/39543 Detección y análisis de datos sobre especies exóticas en biomas diferenciados en apoyo a la biodiversidad empleando la herramienta YOLO y microcontroladores
Loading...
Identifiers
Publication date
Authors
Jiménez Gómez, Xabier
Advisors
Other responsabilities
Universidade da Coruña. Facultade de Informática
Journal Title
Bibliographic citation
Type of academic work
Academic degree
Abstract
[Resumen]: En la actualidad, se ha hablado mucho sobre las nuevas tecnologías que están transformando la sociedad y la forma en que vivimos. Entre estas tecnologías se encuentran el Internet de las Cosas, el Big Data, el Cloud Computing y la Inteligencia Artificial. Es fundamental comprender y explorar su potencial para anticipar cómo impactarán en nuestro futuro. En este proyecto, nos centraremos en una de estas tecnologías: la detección de objetos, una rama de la Inteligencia Artificial que busca permitir a las máquinas ver y comprender el mundo de la misma manera que lo hacemos nosotros a través de nuestros ojos. Nuestro objetivo principal es utilizar la detección de objetos para identificar especies en peligro de extinción o que habitan en biomas de difícil estudio a partir de imágenes de video. Este enfoque tiene un propósito claro: contribuir a la preservación de la biodiversidad, pues el legado que debemos dejar a nuevas generaciones. Implementaremos este sistema en un dispositivo económico y versátil, la Raspberry Pi , que originalmente no fue diseñada con este propósito, para evaluar su rendimiento y determinar su viabilidad. En particular, utilizaremos una versión simplificada de YOLO, uno de los algoritmos más avanzados en detección de objetos en los últimos años, conocido como TinyYOLO. Nuestros objetivos incluyen la instalación y puesta en marcha exitosa del algoritmo, así como la evaluación de su desempeño en aspectos clave de la inteligencia artificial, como el entrenamiento y la inferencia. A lo largo de esta memoria, exploraremos los fundamentos que rigen el campo de la detección de objetos, desde los principios básicos de la inteligencia artificial hasta el deep learning. Luego, detallaremos el proceso de integración del algoritmo TinyYOLO en la Raspberry Pi. Finalmente, llevaremos a cabo pruebas exhaustivas y análisis de datos que nos permitirán alcanzar las conclusiones y objetivos establecidos en este proyecto, todo con el propósito de avanzar en la conservación de la biodiversidad. En este proyecto, no solo nos limitaremos a implementar la detección de objetos en un dispositivo como la Raspberry Pi , sino que también crearemos un dataset de imágenes propio. Este conjunto de datos contendrá imágenes de las especies que estamos estudiando y en las que hemos realizado detecciones exitosas. Nos enfocaremos en identificar estas especies y recopilaremos datos valiosos sobre su comportamiento y hábitats. A medida que avancemos en nuestro trabajo, compartiremos estos datos junto con las detecciones realizadas. Esto enriquecerá nuestra comprensión de estas especies, además de que también proporcionará información valiosa que puede ser utilizada en la preservación y conservación de la biodiversidad. En línea con la filosofía de Michael Jordan, quien dijo: ”He fallado una y otra vez en mi vida. Y es por eso que he tenido éxito”, consideramos que aprender de nuestros errores y esfuerzos en este proyecto nos llevará a lograr avances significativos en la protección de las especies en peligro y la comprensión de los ecosistemas vulnerables. El camino hacia la conservación de la biodiversidad comienza con la recopilación de datos precisos y la implementación de tecnologías innovadoras, y eso es precisamente lo que buscamos lograr aquí.
[Abstract]: Currently, there has been a lot of talk about new technologies that are transforming society and the way we live. Among these technologies are the Internet of Things, Big Data, Cloud Computing and Artificial Intelligence. It is critical to understand and explore their potential in order to anticipate how they will impact our future. In this project, we will focus on one of these technologies: object detection, a branch of Artificial Intelligence that seeks to enable machines to see and understand the world in the same way we do through our eyes. Our main goal is to use object detection to identify endangered species or species that inhabit difficult-to-study biomes from video images. This approach has a clear purpose: to contribute to the preservation of biodiversity, as the legacy we must leave to new generations. We will implement this system on an inexpensive and versatile device, the Raspberry Pi , which was not originally designed for this purpose, to evaluate its performance and determine its feasibility, in particular, we will use a simplified version of YOLO, one of the most advanced algorithms in object detection in recent years, known as TinyYOLO. Our objectives include the successful installation and implementation of the algorithm, as well as the evaluation of its performance in key aspects of artificial intelligence, such as training and inference. Throughout this report, we will explore the fundamentals that govern the field of object detection, from the basics of artificial intelligence to deep learning. Then, we will detail the process of integrating the TinyYOLO algorithm on the Raspberry Pi. Finally, we will carry out extensive testing and data analysis that will allow us to reach the conclusions and objectives established in this project, all with the purpose of advancing biodiversity conservation. In this project, we will not only limit ourselves to implementing object detection on a device such as the Raspberry Pi , but we will also create an image dataset of our own. This dataset will contain images of the species we are studying and on which we have made successful detections. Not only will we focus on identifying these species, but we will also collect valuable data on their behavior and habitats. As we progress in our work, we will share this data along with the detections made. This will not only enrich our understanding of these species, but will also provide valuable information that can be used in the preservation and conservation of biodiversity. In line with the philosophy of Michael Jordan, who said, ”I have failed over and over again in my life. And that is why I have succeeded”, we believe that learning from our mistakes and efforts in this project will lead to significant advances in the protection of endangered species and the understanding of vulnerable ecosystems. The road to biodiversity conservation begins with accurate data collection and the implementation of innovative technologies, and that is precisely what we seek to achieve here.
[Abstract]: Currently, there has been a lot of talk about new technologies that are transforming society and the way we live. Among these technologies are the Internet of Things, Big Data, Cloud Computing and Artificial Intelligence. It is critical to understand and explore their potential in order to anticipate how they will impact our future. In this project, we will focus on one of these technologies: object detection, a branch of Artificial Intelligence that seeks to enable machines to see and understand the world in the same way we do through our eyes. Our main goal is to use object detection to identify endangered species or species that inhabit difficult-to-study biomes from video images. This approach has a clear purpose: to contribute to the preservation of biodiversity, as the legacy we must leave to new generations. We will implement this system on an inexpensive and versatile device, the Raspberry Pi , which was not originally designed for this purpose, to evaluate its performance and determine its feasibility, in particular, we will use a simplified version of YOLO, one of the most advanced algorithms in object detection in recent years, known as TinyYOLO. Our objectives include the successful installation and implementation of the algorithm, as well as the evaluation of its performance in key aspects of artificial intelligence, such as training and inference. Throughout this report, we will explore the fundamentals that govern the field of object detection, from the basics of artificial intelligence to deep learning. Then, we will detail the process of integrating the TinyYOLO algorithm on the Raspberry Pi. Finally, we will carry out extensive testing and data analysis that will allow us to reach the conclusions and objectives established in this project, all with the purpose of advancing biodiversity conservation. In this project, we will not only limit ourselves to implementing object detection on a device such as the Raspberry Pi , but we will also create an image dataset of our own. This dataset will contain images of the species we are studying and on which we have made successful detections. Not only will we focus on identifying these species, but we will also collect valuable data on their behavior and habitats. As we progress in our work, we will share this data along with the detections made. This will not only enrich our understanding of these species, but will also provide valuable information that can be used in the preservation and conservation of biodiversity. In line with the philosophy of Michael Jordan, who said, ”I have failed over and over again in my life. And that is why I have succeeded”, we believe that learning from our mistakes and efforts in this project will lead to significant advances in the protection of endangered species and the understanding of vulnerable ecosystems. The road to biodiversity conservation begins with accurate data collection and the implementation of innovative technologies, and that is precisely what we seek to achieve here.
Description
Editor version
Rights
Atribución 3.0 España







