Web Scraping Methodologies for Image Dataset Creation: A motorcycle Imagery Case Study

Artem Gubin

Bachelor's thesis

Abstract:

This thesis investigates the methods and techniques of web scraping and subsequently applies them in the creation of a structured dataset of images for future computer vision tasks. The work is divided into theoretical and practical parts. The theoretical part describes the objectives of web scraping and provides a thorough analysis of its tools and techniques. Furthermore, it addresses the current challenges of web scraping, including its technical, ethical, and legal aspects, and provides an in-depth discussion of these issues. The practical part demonstrates the application of these methodologies, supported by tools like Selenium and Beautiful Soup, for collecting and processing an image dataset of motorcycles. This process involves developing scrapers, integrating them into a general scraping system, scaling the system, and using neural network filtering to ensure data quality. The final compiled dataset, categorized by manufacturer, model, and years of production, showcases the practical use of web scraping tools for Vehicle Make and Model Recognition (VMMR) tasks. Nevertheless, in summary, the thesis emphasizes the significant potential of employing web scraping methods for the creation of datasets in the field of computer vision tasks.

Keywords

Image Dataset, Vehicle Make and Model Recognition (VMMR), Motorcycle dataset, Selenium, Beautiful Soup, YOLO, Data Filtration, Web Scraping

Jazyk práce: English

Datum vytvoření / odevzdání či podání práce: 18. 3. 2024

Obhajoba závěrečné práce

• Obhajoba proběhla 21. 5. 2024
• Vedúci: Ing. Martin Pelikán, Ph.D.
• Oponent: Jiří Orság, Ing.

Czech University of Life Sciences Prague

Faculty of Economics and Management

Bachelor programme:
Informatics