Related Experiment Video
Updated: Aug 3, 2025
07:34
Probing the Limits of Egg Recognition Using Egg Rejection Experiments Along Phenotypic Gradients
Published on: August 22, 2018
8.3K
Evaluation of Duck Egg Hatching Characteristics with a Lightweight Multi-Target Detection Method.
Jiaxin Zhou1,2, Youfu Liu1,2, Shengjie Zhou1,2
1College of Mathematics Informatics, South China Agricultural University, Guangzhou 510225, China.
Animals : an Open Access Journal From MDPI
|April 13, 2023
Summary
A new lightweight detection architecture (LDA) accurately identifies sterile duck eggs using YOLOX-Tiny. This method improves detection accuracy and efficiency for breeder duck eggs on hatching trays.
Area of Science:
- Agricultural Engineering
- Computer Vision
- Machine Learning
Background:
- Accurate identification of duck egg fertility is challenging.
- Existing detection models lack ease of deployment.
- Need for efficient sterile duck egg detection on incubation trays.
Purpose of the Study:
- Propose a novel lightweight detection architecture (LDA).
- Improve detection accuracy and reduce model deployment requirements for sterile duck eggs.
- Enhance efficiency for single-step detection of breeder duck eggs.
Main Methods:
- Utilized YOLOX-Tiny framework with modifications.
- Implemented depth-wise separable convolution and a new CSP structure.
- Incorporated an attention mechanism and cosine annealing algorithm for training.
- Augmented dataset using rotation, symmetry, and contrast enhancement.
Main Results:
- Achieved 99.74% mean average precision (mAP) on a test set.
- Reduced model parameters to 1.93 M (compared to 5.03 M).
- Demonstrated concurrent detection of 63 eggs in a 7x9 grid.
Conclusions:
- The LDA significantly enhances detection accuracy and efficiency for breeder duck eggs.
- The reduced network size makes it suitable for deployment on hatching egg trays.
- The method offers a practical solution for identifying sterile duck eggs.
