CBL - Campus del Baix Llobregat

Projecte llegit

Títol: Boletify: a machine learning-based application to identify mushrooms

Estudiants que han llegit aquest projecte:

Director/a: OLLER ARCAS, TONI

Departament: ENTEL

Títol: Boletify: a machine learning-based application to identify mushrooms

Data inici oferta: 01-03-2021     Data finalització oferta: 01-11-2021


Estudis d'assignació del projecte:
    GR ENG SIS TELECOMUN
    GR ENG SIST AEROESP
    GR ENG TELEMÀTICA
Tipus: Individual
 
Lloc de realització: EETAC
 
Paraules clau:
Machine Learning, Tensorflow Lite, Flutter, Firebase, backend as a service, cross-platform frameworks, mushrooms, classifiers, release
 
Descripció del contingut i pla d'activitats:
 
Overview (resum en anglès):
Machine learning-based solutions have become very popular in recent years. Although it has been decades since its first use, the technology has gained popularity and adoption thanks to tools such as Jupyter or Anaconda, as well as the increase in the research on neural networks, deep learning and artificial intelligence. Examples of this could be automatic driving, classifiers or data analysis.

The presented applications allow classifying mushrooms by image recognition through a classifier model implemented with Tensorflow Lite, which is specially designed to work on mobile and edge devices. Furthermore, as the classification would usually be used in places where there is no internet connection we have developed an offline mode that keeps all functionalities in those cases. We have been able to achieve it by moving the classifier model inside the native applications instead of getting the response from the server, and also by creating a copy of the mushroom information in the mobile¿s local storage.

The frontend of the application has been developed with Flutter, a relatively booming framework that allows us to create applications for Android, iOS, web, and desktop using almost the same code.

The outcome is an Android app in Alpha version that is available from the Play Store only for internal testers, it is able to identify several types of mushrooms and to work offline. The publishing and integration process has been automated, so changes made in the Github repository generate a new version which is uploaded directly to the Play Store.

© CBLTIC Campus del Baix Llobregat - UPC