In logistics and warehouse management, accurate item counting is a critical task that can benefit significantly from automation through computer vision. This thesis investigates the use of object detection models for the automated counting of books stacked on pallets. The goal is to evaluate and compare the performance of three different deep learning-based object detection frameworks: Dave, YOLO, and Detectron2. Each model receives as input an image of palletized books and outputs a count based on the detected objects. The analysis includes performance metrics such as counting accuracy, inference time, and robustness to variations in image conditions. The results highlight the trade-offs between accuracy and computational efficiency across the models and provide insights into their suitability for real-world industrial applications. This work aims to contribute to the development of intelligent visual systems for logistics automation.
Nel contesto della logistica e della gestione dei magazzini, il conteggio accurato degli oggetti rappresenta un'attività cruciale che può essere notevolmente migliorata tramite l'automazione basata sulla visione artificiale. Questa tesi analizza l’utilizzo di modelli di object detection per il conteggio automatico dei libri presenti sui pallet. L’obiettivo è valutare e confrontare le prestazioni di tre diversi modelli di deep learning: Dave, YOLO e Detectron2. Ciascun modello riceve in input un’immagine contenente libri palletizzati e restituisce in output un conteggio basato sugli oggetti rilevati. L’analisi include metriche di performance come l’accuratezza del conteggio, i tempi di inferenza e la robustezza rispetto a variazioni nelle condizioni dell’immagine. I risultati evidenziano i compromessi tra accuratezza ed efficienza computazionale, offrendo indicazioni utili sulla loro applicabilità in scenari industriali reali. Questo lavoro mira a contribuire allo sviluppo di sistemi visivi intelligenti per l’automazione logistica.
Conteggio automatizzato di libri in ambienti logistici tramite reti di Object Detection: Dave, YOLO e Detectron2.
MARINO, DAVIDE
2024/2025
Abstract
In logistics and warehouse management, accurate item counting is a critical task that can benefit significantly from automation through computer vision. This thesis investigates the use of object detection models for the automated counting of books stacked on pallets. The goal is to evaluate and compare the performance of three different deep learning-based object detection frameworks: Dave, YOLO, and Detectron2. Each model receives as input an image of palletized books and outputs a count based on the detected objects. The analysis includes performance metrics such as counting accuracy, inference time, and robustness to variations in image conditions. The results highlight the trade-offs between accuracy and computational efficiency across the models and provide insights into their suitability for real-world industrial applications. This work aims to contribute to the development of intelligent visual systems for logistics automation.| File | Dimensione | Formato | |
|---|---|---|---|
|
Marino_Davide.pdf
embargo fino al 16/10/2028
Dimensione
32.05 MB
Formato
Adobe PDF
|
32.05 MB | Adobe PDF |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/94402