My Eye AI: Object Detection System Using YOLO, OWL-ViT, and BLIP

This paper presents My EYE AI, an innovative technology application that empowers visually impaired users by providing real-time object detection and comprehensive scene description. The system employs YOLOv11 for object detection, OWL- ViT for zero-shot recognition, and BLIP for natural language scene description. In addition, two variants (medium and X-large) of YOLOvll are used to retrain the model using a diverse dataset of household objects. The training results are analyzed, highlighting key innovations in the dataset reannotation, hyperparameter tuning, and model evaluation. Experimental results demonstrate promising performance improvement while identifying class imbalance and overfitting challenges. Future work will address these issues and optimize the system for efficient edge development.