Related Experiment Video
Updated: Sep 1, 2025
07:43
Quail Chorioallantoic Membrane - A Tool for Photodynamic Diagnosis and Therapy
Published on: April 28, 2022
2.9K
Development of an Early Embryo Detection Methodology for Quail Eggs Using a Thermal Micro Camera and the YOLO Deep
Victor Massaki Nakaguchi1, Tofael Ahamed2
1Graduate School of Science and Technology, University of Tsukuba, Tennodai 1-1-1, Tsukuba 305-8577, Japan.
Sensors (Basel, Switzerland)
|August 12, 2022
Summary
Thermal imaging and YOLO deep learning can detect quail embryos early in incubation. This technology offers a cost-effective solution for poultry production, improving embryo detection and potentially aiding sex segregation.
Area of Science:
- Agricultural Engineering
- Animal Science
- Computer Vision
Background:
- Poultry production faces challenges with early-stage avian embryo detection and sex segregation due to a lack of affordable, robust technologies.
- Current methods for embryo monitoring in poultry are often labor-intensive or lack precision in the critical early incubation phases.
Purpose of the Study:
- To evaluate the efficacy of thermal micro-cameras for detecting quail embryos within the first 168 hours of incubation.
- To apply deep learning object detection algorithms (YOLO, SSD-MobileNet V2) for distinguishing fertilized from unfertilized quail eggs.
- To investigate the impact of egg turning frequency on the accuracy of thermal imaging-based embryo detection.
Main Methods:
- Collected thermal images of quail eggs during the initial 7 days of incubation.
- Trained and compared YOLOv4, YOLOv5, and SSD-MobileNet V2 models for object detection of embryos.
- Assessed model performance using mean Average Precision (mAP@0.50) and F1 Scores across different egg-turning intervals (1.5h, 6h, 12h).
Main Results:
- YOLOv5 achieved the highest mAP@0.50 (99.5%), followed by YOLOv4 (98.62%) and SSD-MobileNet V2 (91.8%) in detecting fertilized eggs.
- YOLOv5 demonstrated perfect F1 Scores (1.0) for embryo detection at a 12-hour turning interval, significantly outperforming other models and intervals.
- Egg turning frequency influenced detection accuracy, with less frequent turning (12h) generally yielding better results for some models, though not linearly.
Conclusions:
- Thermal imaging combined with YOLO deep learning presents a promising, cost-effective method for early avian embryo detection in quail eggs.
- The YOLOv5 algorithm shows exceptional potential for high-accuracy embryo detection, particularly when combined with optimized egg-turning strategies.
- This technology could significantly advance farm-industry automation and sanitary control in poultry production by enabling early and accurate embryo assessment.
