Study on an Enhanced Method for Riparian Waste Detection Using Drone Images and a Large-Scale Multimodal Model

Key Points

• The research aims to improve the detection and classification of river waste using drone images and advanced AI models.
• Utilized UAV aerial photography for data collection.
• Employed the YOLO object detection model for identifying waste.
• Used the LLaVA multimodal model for classifying detected waste.
• Analyzed the impact of capturing height and AI-generated images on detection accuracy.
• Detection accuracy improved when using low-height aerial images.
• YOLO detected waste effectively, while LLaVA accurately classified larger items like Styrofoam.
• Higher capture heights and smaller objects resulted in decreased classification accuracy.