UAV weed detection models balance accuracy and speed for edge devices

Researchers have developed a framework for deploying weed detection models on resource-constrained UAVs for site-specific management. The study evaluated various object detection models, including YOLO and RT-DETR variants, across different edge devices like Jetson Orin Nano and Jetson AGX Xavier. Results indicated a trade-off between detection accuracy and computational efficiency, with high-capacity models achieving better accuracy but slower inference times. Lightweight models offered real-time performance, and RT-DETRv2-R50-M and YOLOv11s emerged as strong candidates for balancing accuracy and speed in practical UAV applications. AI

IMPACT Provides insights into optimizing AI model deployment for real-time edge applications in agriculture.

RANK_REASON This is a research paper evaluating existing models for a specific application.

Read on arXiv cs.CV →

• Jetson AGX Xavier
• Jetson AGX Orin
• Jetson Orin Nano
• RT-DETR
• UAV
• YOLO

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →