Related Experiment Video
Updated: Jul 10, 2025
11:02
The Use of an Automated System GreenFeed to Monitor Enteric Methane and Carbon Dioxide Emissions from Ruminant Animals
Published on: September 7, 2015
22.0K
Improving Known-Unknown Cattle's Face Recognition for Smart Livestock Farm Management.
Yao Meng1,2, Sook Yoon3, Shujie Han1,2
1Department of Electronic Engineering, Jeonbuk National University, Jeonju 54896, Republic of Korea.
Animals : an Open Access Journal From MDPI
|November 25, 2023
Summary
This study introduces a new method for identifying individual cattle faces, even unknown ones, improving precision livestock farming. The approach enhances accuracy in recognizing both known and novel cattle, aiding herd management.
Area of Science:
- Precision Livestock Farming
- Computer Vision
- Animal Science
Background:
- Accurate individual cattle identification is crucial for precision livestock farming, disease prevention, and animal welfare.
- Distinguishing individual Hanwoo cattle is challenging due to facial similarities and uniform body color.
- Existing methods often struggle with open-set recognition, where unknown individuals must also be identified.
Purpose of the Study:
- To develop a robust Cattle's Face Open-Set Recognition (CFOSR) system capable of identifying both known and unknown Hanwoo cattle.
- To enhance closed-set identification accuracy by integrating open-set recognition techniques.
- To improve herd monitoring and inventory management through reliable individual cattle identification.
Main Methods:
- Integration of Adversarial Reciprocal Points Learning (ARPL) to reduce feature space overlap between known and unknown cattle.
- Application of Additive Margin Softmax loss (AM-Softmax) for improved classification of known individuals, outperforming conventional Cross-Entropy loss.
- Validation using a real-world dataset to assess performance in both open-set and closed-set recognition scenarios.
Main Results:
- The proposed CFOSR method demonstrated superior performance in open-set recognition with an AUROC of 91.84 and OSCR of 87.85.
- Closed-set recognition accuracy reached 94.46%, validating the effectiveness of the integrated techniques.
- Empirical results confirmed enhanced intra-class compactness and inter-class separability, crucial for accurate identification.
Conclusions:
- The novel CFOSR approach effectively addresses the challenge of identifying known and unknown cattle in precision livestock farming.
- The integration of ARPL and AM-Softmax offers a significant advancement over existing algorithms for cattle face recognition.
- This study provides a foundation for improved herd management systems, particularly in diverse or dynamic herd environments.
