Sign in Subscribe
Machine learning

Breakthrough in Tomato Monitoring: Improved YOLOv8 Model for Robust Flower Bud Detection

Strategic Insights

— 1 min read

Research conducted at the Beijing Academy of Agricultural and Forestry Sciences has made a significant breakthrough in the development of an improved YOLOv8 model for robust flower bud detection in complex facility environments. The study aimed to address the challenges of detecting vital physiological indicators such as tomato growing points and flower buds, which significantly influence yield quality. The researchers constructed a comprehensive multi-environment dataset covering 10 typical growing conditions with enhanced annotations to improve the detection accuracy.

Key Takeaways:

  • The improved YOLOv8 model addresses the limitations of the original model by incorporating three key innovations: an SE attention module, GhostConv, and a scale-adaptive WIoU_v2 loss function.
  • These modifications synergistically improve adaptability to scale and environmental variations, achieving 97.8% mAP@0.5 (+ 0.5%) and 85.1% mAP@0.5:0.95 (+ 5.1%) with 11% fewer parameters.
  • The proposed system achieves an optimal balance of accuracy, speed, and lightweight design while providing immediately applicable solutions for automated tomato monitoring.
  • The model was successfully deployed on agricultural robots in operational greenhouses, demonstrating 93.6% detection accuracy.
  • The research emphasizes the importance of accurate and efficient detection of flower buds and growing points for precision agriculture applications.
  • The study was conducted by a team of researchers from the Beijing Academy of Agricultural and Forestry Sciences, including Jingxin Yu, Jiang Liu, Changfu Zhang, Huankang Cui, Jinpeng Zhao, Wengang Zheng, Fan Xu, and Xiaoming Wei.

Statistics:

  • 97.8% mAP@0.5 (+ 0.5%) detection accuracy achieved by the improved YOLOv8 model
  • 85.1% mAP@0.5:0.95 (+ 5.1%) detection accuracy achieved by the improved YOLOv8 model
  • 11% reduction in parameters while maintaining detection accuracy
  • 93.6% detection accuracy achieved by the proposed system in operational greenhouses

Sources:

  • Optimization of a multi-environmental detection model for tomato growth point buds based on multi-strategy improved YOLOv8. Scientific Reports, 2025;15(1):25726.
  • Scientific Reports can be contacted at: Nature Portfolio, Heidelberger Platz 3, Berlin, 14197, Germany.
  • NewsRx. Beijing Academy of Agricultural and Forestry Sciences Reports Findings in Robotics and Machine Learning (Optimization of a multi-environmental detection model for tomato growth point buds based on multi-strategy improved YOLOv8). Journal of Engineering. July 28, 2025; p 153.

Read more