Small Object Detection with integration of Celeste

This project focuses on building a robust robot vision pipeline for object recognition and pose estimation using deep learning. It involves automated dataset generation and augmentation, including rotation-aware transformations, to train YOLO-based oriented bounding box (OBB) models.
The system is designed to handle real-world visual variability and support reliable perception for robotic manipulation tasks, bridging data engineering, computer vision, and applied AI for physical systems.