Improving the Recognition of Bamboo Color and Spots Using a Novel YOLO Model

Area of Science:

• Agricultural Science
• Computer Science
• Genetics

Background:

• Bamboo shoot sheath coloration and spotting are crucial phenotypic markers for species identification, economic valuation, and genetic research.
• Accurate and efficient methods for analyzing these traits are essential for bamboo germplasm evaluation and conservation efforts.

Purpose of the Study:

• To develop and evaluate a deep learning model, YOLOv8-BS, for the precise detection of color and spot patterns on bamboo shoot sheaths.
• To compare the performance of YOLOv8-BS against established object detection models for phenotypic trait analysis in *Chimonobambusa utilis*.

Main Methods:

• Utilized a dataset of *Chimonobambusa utilis* shoot sheaths from Jinfo Mountain, China.
• Applied data augmentation techniques including translation, flipping, and contrast adjustment to enhance the training dataset.
• Implemented and benchmarked the YOLOv8-BS model against YOLOv7, YOLOv5, YOLOX, and Faster R-CNN for color and spot detection.

Main Results:

• YOLOv8-BS demonstrated superior performance in detecting both color and spot patterns compared to benchmark models.
• For color detection, YOLOv8-BS achieved a precision of 85.9%, recall of 83.4%, F1-score of 84.6%, and an average precision (AP) of 86.8%.
• For spot detection, YOLOv8-BS achieved a precision of 90.1%, recall of 92.5%, F1-score of 91.1%, and an AP of 96.1%.

Conclusions:

• The YOLOv8-BS model offers a highly accurate and robust solution for automated phenotypic analysis of bamboo shoot sheaths.
• This deep learning approach facilitates precise germplasm evaluation, genetic diversity studies, and supports the development of sustainable bamboo-based industries through precision agriculture.
• Future work may focus on enhancing the model for fine-grained varietal distinctions and real-time application capabilities.