Related Experiment Video
Updated: Jul 8, 2025
06:03
AMEBaS: Automatic Midline Extraction and Background Subtraction of Ratiometric Fluorescence Time-Lapses of Polarized Single Cells
Published on: June 23, 2023
493
AC-UNet: an improved UNet-based method for stem and leaf segmentation in Betula luminifera
Xiaomei Yi1, Jiaoping Wang1, Peng Wu1
1College of Mathematics and Computer Science, Zhejiang A&F University, Hangzhou, China.
Frontiers in Plant Science
|December 13, 2023
Summary
This study introduces AC-UNet, an improved algorithm for segmenting plant stems and leaves, enhancing accuracy in plant phenotyping for genetic breeding. The new method achieves superior performance over existing models.
Area of Science:
- Plant Science
- Computer Vision
- Bioinformatics
Background:
- Accurate plant phenotyping is crucial for understanding plant growth and genetic traits.
- Automated segmentation of plant organs like stems and leaves aids in efficient growth monitoring and large-scale genetic breeding.
Purpose of the Study:
- To develop an improved UNet-based algorithm (AC-UNet) for precise stem and leaf segmentation in Betula luminifera.
- To address challenges in feature edge information loss and sample breakage during plant organ segmentation.
Main Methods:
- The AC-UNet algorithm utilizes VGG16 as its backbone for feature extraction, incorporating a multi-scale mechanism and an optimized pyramid pooling module.
- A cross-attention mechanism is added to the expanding network, and Dice_Boundary loss function is employed to handle sample imbalance.
Main Results:
- AC-UNet achieved a mean Intersection over Union (mIoU) of 87.50%, mean PA (mPA) of 92.71%, and Precision of 93.69% on the dataset.
- The proposed algorithm outperformed PSPNet, DeepLabV3, traditional UNet, and Swin-UNet in segmentation accuracy and speed.
- AC-UNet demonstrated efficient segmentation of Betula luminifera stems and leaves.
Conclusions:
- The AC-UNet algorithm provides a highly accurate and efficient solution for plant organ segmentation.
- This method offers improved auxiliary support for acquiring plant phenotypic traits, benefiting genetic research and breeding programs.
