Related Experiment Video
Updated: May 13, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.6K
Transfer learning-based approach to individual Apis cerana segmentation
Panadda Kongsilp1, Unchalisa Taetragool1, Orawan Duangphakdee2
1Department of Computer Engineering, King Mongkut's University of Technology Thonburi, Bangkok, Thailand.
Plos One
|April 16, 2025
Summary
This study introduces an efficient method for detecting and segmenting Asian honey bees (Apis cerana) using Mask R-CNN. The approach significantly reduces data requirements while achieving high accuracy in bee segmentation.
Area of Science:
- Ecology
- Computer Science
- Zoology
Background:
- Honey bees are vital pollinators, exhibiting complex social structures and communication within hives.
- Automated analysis of honey bee behavior relies on accurate individual bee detection and segmentation.
- Understanding Apis cerana behavior is crucial for ecological and agricultural applications.
Purpose of the Study:
- To develop an accurate and efficient method for detecting and segmenting individual Apis cerana bees within a hive.
- To optimize the use of deep learning models for bee behavior analysis with minimal data and computational resources.
Main Methods:
- Utilized the Mask R-CNN deep learning model for object detection and segmentation.
- Employed transfer learning from a pre-existing Apis mellifera model.
- Applied data preprocessing techniques, including brightness and contrast enhancement.
- Evaluated performance using mean average precision (mAP) for both detection and segmentation.
Main Results:
- Achieved a mean average precision (mAP) of 0.728 for Apis cerana segmentation.
- Reduced the size of training and validation datasets by 85% compared to previous studies.
- Demonstrated high model performance with a minimal dataset size and reduced computational time.
Conclusions:
- The proposed Mask R-CNN approach provides an optimal solution for Apis cerana detection and segmentation.
- This method significantly enhances automated behavior analysis by enabling efficient, high-quality individual bee tracking.
- The study highlights the effectiveness of transfer learning and data preprocessing for improving deep learning model performance in ecological contexts.
