Related Experiment Video
Updated: Jun 30, 2025
06:41
Imaging and Analysis for Quantifying Maize (Zea mays) Abiotic Stress Phenotypes
Published on: March 28, 2025
799
Advanced deep learning models for phenotypic trait extraction and cultivar classification in lychee using
Mengjia Xue1, Siyi Huang1, Wenting Xu2
1Institute of Radiation Medicine, Fudan University, Shanghai, China.
Frontiers in Plant Science
|March 15, 2024
Summary
Advanced photon-counting micro-CT and deep learning models precisely characterize lychee fruit phenotypes. This non-destructive imaging approach significantly improves the identification and classification of lychee varieties.
Area of Science:
- Agronomic research
- Plant phenotyping
- Non-destructive imaging
Background:
- Contemporary agronomic research emphasizes non-destructive imaging and precise phenotypic characterization.
- A novel photon-counting micro-CT system with advanced detectors enables micrometer-level imaging and full energy spectrum capture of lychee fruit.
Purpose of the Study:
- To develop and apply deep learning models for automatic measurement of phenotypic traits in lychee fruit.
- To utilize machine learning for identifying key characteristics and classifying different lychee species.
Main Methods:
- Seven Convolutional Neural Network (CNN)-based deep learning models (AttentionUNet, DeeplabV3+, SegNet, TransUNet, UNet, UNet++, UNet3+) were developed.
- Machine learning techniques suitable for small-sample training were employed for feature selection and classification.
Main Results:
- The developed models achieved high performance, with Dice, Recall, and Precision indices ranging from 0.90 to 0.99.
- Mean Intersection over Union (MIoU) values consistently fell between 0.88 and 0.98.
- The approach effectively enabled feature selection and classification, significantly improving the discernment of lychee varieties.
Conclusions:
- This research advances non-destructive plant analysis techniques.
- The study opens new avenues for exploring intricate phenotypic variations within plant species, particularly lychee.
