Related Experiment Video
Updated: May 11, 2025
11:49
Cereal Crop Ear Counting in Field Conditions Using Zenithal RGB Images
Published on: February 2, 2019
9.2K
A method of rice panicle number counting based on improved CSRNet model
Junyan Zhu1,2,3, Jiaxing Huang4, Yiyuan Wang2,3
1College of Agriculture, Fujian Agriculture and Forestry University, Fuzhou, Fujian, China.
Frontiers in Plant Science
|April 18, 2025
Summary
This study developed an improved deep learning model to accurately count rice grains per panicle using smartphone images. The method achieved a 3.83% mean error, enabling practical applications for rice yield estimation.
Area of Science:
- Agricultural Science
- Computer Vision
- Deep Learning
Background:
- Rice is a global staple food, with yield heavily influenced by grain count per panicle.
- Accurate estimation of rice grains per panicle is crucial for yield assessment and breeding programs.
Purpose of the Study:
- To develop and validate an improved deep learning model for automated rice grain counting per panicle.
- To create user-friendly software applications for real-time and batch counting of rice grains.
Main Methods:
- Utilized smartphone-captured images of rice panicles.
- Applied an improved Convolutional Siamese Residual Network (CSRNet) model for grain counting.
- Developed an Android application and PC software for practical deployment.
Main Results:
- The improved CSRNet model achieved a mean error of 3.83% on the validation set.
- Developed a real-time counting Android APP and batch counting PC software (RiceGrainCounter).
- Demonstrated the model's effectiveness for theoretical and technical support in rice per panicle counting.
Conclusions:
- The deep learning-based approach offers a highly accurate and efficient method for counting rice grains per panicle.
- The developed applications provide practical tools for farmers and researchers to support rice breeding and yield estimation.
- This technology can significantly aid in optimizing rice production and management strategies.
