Percorrer por autor "Prata, Filipe Miguel Maia"
A mostrar 1 - 1 de 1
Resultados por página
Opções de ordenação
- Enhancing moving services with AI: a Deep Learning approach for furniture detectionPublication . Prata, Filipe Miguel Maia; Ramos, Carlos Fernando da SilvaThe moving services industry, traditionally reliant on manual processes and estimations, faces significant challenges in accurately identifying and estimating the volume of furniture for service quotations. This thesis explores the integration of advanced AI technologies to address these challenges, focusing on the development of a mobile application that leverages artificial intelligence for furniture detection and volume estimation for moving services. Machine Learning, particularly Deep Learning, has demonstrated considerable success in tackling complex problems such as object detection and image recognition. This thesis capitalizes on these advancements by initially employing traditional object detection models to classify and estimate the volume of furniture from static images. However, these models, while effective within certain constraints, were limited by their inability to accurately differentiate between overlapping classes, detect sizes, and handle complex furniture configurations. Recognizing these limitations, the research pivoted to integrate GPT-4o, a state-of-the-art multimodal AI model, which brought significant improvements in detection accuracy and contextual understanding. Alongside the development of this application, a thorough study was conducted on the evolution and effectiveness of different machine learning architectures, with a deep focus on Convolutional Neural Networks (CNNs) and their advancements in object detection tasks. This study provided a comprehensive comparison of these architectures, illustrating their strengths and weaknesses in the context of the moving services industry. The integration of GPT-4o into the system allowed for superior performance, particularly in scenarios where traditional models struggled. This enhanced the application's ability to deliver more accurate and reliable service quotations, ultimately improving operational efficiency and customer satisfaction. The thesis concludes by reflecting on the project's achievements, including the successful application of advanced AI models, and suggests avenues for future research, particularly in fine-tuning AI models for specific use cases and exploring new AI technologies as they emerge.