Lane type classification using efficient networks

This thesis presents the result of the research, analysis and development of algorithms based on efficient deep convolutional networks. The goal of the project is that the algorithms must be able to perform a lane type classification task with the aim of making micro-mobility vehicles (such as bicyc...

Descripción completa

Detalles Bibliográficos
Autor: Mateo I Remacha, Alex
Tipo de recurso: tesis de maestría
Fecha de publicación:2022
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/379857
Acceso en línea:https://hdl.handle.net/2117/379857
Access Level:acceso abierto
Palabra clave:Image analysis
Deep learning (Machine learning)
Image classification
Deep learning
Efficient Networks
Imatges--Anàlisi
Aprenentatge profund
Àrees temàtiques de la UPC::Enginyeria de la telecomunicació::Processament del senyal::Processament de la imatge i del senyal vídeo
Descripción
Sumario:This thesis presents the result of the research, analysis and development of algorithms based on efficient deep convolutional networks. The goal of the project is that the algorithms must be able to perform a lane type classification task with the aim of making micro-mobility vehicles (such as bicycles or scooters) comply with the speed limits according to the type of road in order to improve the safety of both people and drivers. To achieve this goal, a set of different efficient networks for image classification have been trained and tested to decide which one is the most suitable for this project. In addition to the performance of the model, both inference time and size have been taken into account since the idea is to perform the image classification task from a mobile phone, so the requirements are quite strict.