PORTALE DELLA DIDATTICA

Ricerca CERCA
  KEYWORD

Automatic classification of images of food

Parole chiave DEEP NEURAL NETWORKS, IMAGE PROCESSING, MACHINE LEARNING

Riferimenti ANDREA BOTTINO, MAURIZIO MORISIO

Gruppi di ricerca GR-16 - SOFTWARE ENGINEERING GROUP - SOFTENG

Tipo tesi SPERIMENTALE APPLICATA

Descrizione In many contexts it is important to track what a person eats (illnesses, fitness, allergies etc). A practical way to do this is to take a picture of the food eaten using a smartphone and having an automatic classifier capable of recognizing the food with high accuracy. Having recognized the food it is then possible to know the basic ingredients eaten (proteins, fats etc). Beyond food classification another open problem is characterizing the quantity eaten.
This thesis consists in selecting a suitable open data set of tagged food pictures (ex Recipe1M+), trying various machine learning / deep learning techniques to build a classifier, using image processing techniques to infer the quantity of food eaten, with the goal of achieving the highest accuracy both in food classification and quantity computation.

The thesis is part of a research project aiming at using images collected with drones to monitor the status of plants (notably hazelnut trees).
The drone collects images (optical and infrared spectrums) of plants in regular intervals (weeks), The images are analyzed to classify each plant as healthy or non healthy.
In a second phase non healthy plants should be further classified in subclasses (water stress, insect attack, virus attack, etc).
The project team includes an agricultural company, that provides the plants, and a CNR research institute that provides botanical and agronomic know how.
The student will receive a dataset of tagged images, and will focus on finding and tuning machine learning algorithms to classify plants with the highest accuracy.

Conoscenze richieste Python, Java, machine learning approaches and libraries, focusing on image processing, image classification


Scadenza validita proposta 13/10/2023      PROPONI LA TUA CANDIDATURA




© Politecnico di Torino
Corso Duca degli Abruzzi, 24 - 10129 Torino, ITALY
Contatti