Evaluating Object Detection Models for Small Object Recognition

Bugueño Cuadra, Gonzalo Mauricio (2025) Evaluating Object Detection Models for Small Object Recognition. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (746kB) | Preview

Abstract

Performance in object detection is what makes a detector useful. This study aims to explore different techniques to improve object detection performance for challenging tasks, including dataset composition, augmentation techniques and synthetic data generation. For this purpose, a multitude of You Only Look Once (YOLO) models were trained with different datasets (ranging from annotated photographs to fully synthetic, augmented frames) and hyperparameters. The study focuses on the specific challenge of detecting coins in different settings (indoors, street) - a task inspired by the visual similarity between coins and typical urban litter like chewing gum. This scenario amplifies the challenge due to subtle textural differences, ambiguous object boundaries, and warping shape. The study demonstrates that it is possible to build a performant, small object detector with very limited resources by leveraging Blender’s scripting API, applying augmentations, and choosing the right set of parameters which constitute the model. The final results highlight the importance of dataset composition and model architecture.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Raj, Kislay UNSPECIFIED
Subjects:	Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence > Computer vision Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence > Computer vision
Divisions:	School of Computing > Master of Science in Artificial Intelligence
Depositing User:	Ciara O'Brien
Date Deposited:	28 May 2026 13:01
Last Modified:	28 May 2026 13:01
URI:	https://norma.ncirl.ie/id/eprint/9312

Actions (login required)

View Item