Use this link to cite:
http://hdl.handle.net/2183/34022 Identificación de Ratas de Laboratorio mediante Visión Artificial y Aprendizaje Máquina
Loading...
Identifiers
Publication date
Authors
López López, Diego Antonio
Advisors
Other responsabilities
Universidade da Coruña. Facultade de Informática
Journal Title
Bibliographic citation
Type of academic work
Academic degree
Abstract
[Resumen]: En la actualidad, la experimentación animal continúa vigente: en la investigación y
monitorización del comportamiento, la farmacología y la ecotoxicología, en los estudios
de los efectos del cambio climático, en los estudios de procesos de aprendizaje, así
como en la monitorización de granjas. Considerando necesario e importante acompañar
la investigación mediante la experimentación animal, es relevante mencionar los
programas de seguimiento de imágenes. Mediante la determinación de la posición de los
animales, cada imagen, y relacionando las posiciones del animal a lo largo del tiempo
para generar su trayectoria, podemos extraer información de gran utilidad. El problema
reside en relacionar cada posición con el animal que le corresponde cuando estudiamos
múltiples individuos. Los momentos críticos son los instantes de oclusión entre
individuos o con el entorno. En este proyecto estudiaré y validaré las de técnicas de
aprendizaje máquina y transferencia del conocimiento para el estudio de ratas de
laboratorio en imágenes. El aprendizaje máquina son un grupo de técnicas capaces de
aprender relaciones entre datos. La metodología de la transferencia del conocimiento se
basa en la reutilización de técnicas previamente entrenadas en tareas similares.
Mediante la velocidad del entrenamiento del aprendizaje por transferencia, en
contraposición con enfoques tradicionales de aprendizaje máquina, estas técnicas
suponen una optimización sustancial en el tiempo de entrenamiento de los modelos. Así,
el resultado fundamental de este trabajo, es demostrar la viabilidad de las técnicas de
aprendizaje máquina con transferencia del conocimiento. para identificar animales de
laboratorio como las ratas en videos de ensayos animales.
[Abstract]: Today, animal experimentation continues: in behavioural research and monitoring, pharmacology and ecotoxicology, in studies of the effects of climate change, in studies of learning processes, as well as in farm monitoring. Considering it necessary and important to accompany research through animal experimentation, it is relevant to mention image-tracking programs. By determining the position of the animals in each image and relating the animal's positions over time to generate its trajectory, we can extract useful information. The problem arises when relating each position to the corresponding animal when we study multiple individuals. The critical moments occur in the moments of occlusion between individuals or with the environment. In this research, I will study and validate machine learning and knowledge transfer techniques for the study of laboratory rats in images. Machine learning is a group of techniques capable of learning relationships between data. The knowledge transfer methodology based on the reuse of techniques previously trained in similar tasks. Through the speed of transfer learning training, as opposed to traditional machine learning approaches, these techniques provide substantial optimization in model training time. Thus, the fundamental result of this work is to demonstrate the feasibility of machine learning techniques with knowledge transfer to identify laboratory animals such as rats in animal testing videos.
[Abstract]: Today, animal experimentation continues: in behavioural research and monitoring, pharmacology and ecotoxicology, in studies of the effects of climate change, in studies of learning processes, as well as in farm monitoring. Considering it necessary and important to accompany research through animal experimentation, it is relevant to mention image-tracking programs. By determining the position of the animals in each image and relating the animal's positions over time to generate its trajectory, we can extract useful information. The problem arises when relating each position to the corresponding animal when we study multiple individuals. The critical moments occur in the moments of occlusion between individuals or with the environment. In this research, I will study and validate machine learning and knowledge transfer techniques for the study of laboratory rats in images. Machine learning is a group of techniques capable of learning relationships between data. The knowledge transfer methodology based on the reuse of techniques previously trained in similar tasks. Through the speed of transfer learning training, as opposed to traditional machine learning approaches, these techniques provide substantial optimization in model training time. Thus, the fundamental result of this work is to demonstrate the feasibility of machine learning techniques with knowledge transfer to identify laboratory animals such as rats in animal testing videos.
Description
Editor version
Rights
Todos los derechos reservados. All rights reserved.






