Related Experiment Video
Updated: Jul 2, 2026
08:32
Using Computer-based Image Analysis to Improve Quantification of Lung Metastasis in the 4T1 Breast Cancer Model
Published on: October 2, 2020
6.3K
Determining the Presence and Size of Shoulder Lesions in Sows Using Computer Vision
Shubham Bery1, Tami M Brown-Brandl1, Bradley T Jones2
1Department of Biological Systems Engineering, University of Nebraska-Lincoln, Lincoln, NE 68583, USA.
Animals : an Open Access Journal From MDPI
|January 11, 2024
Summary
Computer vision accurately detects and measures shoulder sores in breeding sows. This technology aids in improving animal welfare and reducing economic losses associated with these lesions.
Area of Science:
- Veterinary Medicine
- Computer Science
- Animal Science
Background:
- Shoulder sores are a significant welfare and economic issue in breeding sows, leading to culling.
- Lesion prevalence varies widely (5-50%) based on flooring, body condition, and lameness.
- Current management involves labor-intensive treatment and medication.
Purpose of the Study:
- To evaluate computer vision for detecting and measuring shoulder lesions in sows.
- To assess the efficacy of deep learning models for lesion localization and size estimation.
Main Methods:
- Utilized a Microsoft Kinect V2 camera for top-down depth and RGB imaging.
- Collected 824 RGB images from 70 sows with varying lesion stages.
- Implemented and compared YOLOv5, YOLOv8, and Faster-RCNN for lesion detection.
- Employed traditional and deep learning (U-Net) segmentation for area estimation.
Main Results:
- YOLOv5 achieved the highest detection performance with an mAP@0.5 of 0.92.
- Object detection models successfully localized lesion areas.
- Segmentation techniques enabled lesion size estimation.
Conclusions:
- Computer vision offers a promising, effective method for detecting and assessing sow shoulder lesions.
- This technology can enhance sow welfare and mitigate economic impacts.
- Further development can lead to automated monitoring systems.
