核心概念
Utilizing Faster R-CNN for precise cannabis seed variant detection.
摘要
The study focuses on the importance of accurately identifying cannabis seed variants for agricultural purposes, regulatory compliance, and market demands. By employing a state-of-the-art object detection model, Faster R-CNN, on a dataset of 17 distinct cannabis seed classes in Thailand, the study achieved high performance metrics such as an mAP score of 94.08% and an F1 score of 95.66%. The research aims to enhance precision breeding and improve agricultural productivity by visually identifying different cannabis seed types using deep neural network models.
統計資料
The study achieved a mAP score of 94.08% and an F1 score of 95.66%.
引述
"No single loss function dominates the performance across all classes."
"The mL1 model emerges as the most effective, showcasing superior mAP at various IoU thresholds."
"Real-time inference performance reveals a trade-off between accuracy and speed."