Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Parallel Attention Network for Cattle Face Recognition

Jiayu Li, Xuechao Zou, Shiying Wang, Ben Chen, Junliang Xing, Pin Tao

TL;DR

This work tackles cattle face recognition in wild environments by introducing the ICRWE dataset and a novel PANet architecture. PANet employs a parallel Transformer backbone with two attention pathways, PAM and FMM, to capture local and global features and generate discriminative 128-dimensional embeddings via triplet learning. On ICRWE, PANet achieves a state-of-the-art accuracy of $88.03\%$, outperforming multiple SOTA methods while maintaining efficient model size. The dataset and code availability promise practical impact for wildlife monitoring and precision livestock management in uncontrolled environments.

Abstract

Cattle face recognition holds paramount significance in domains such as animal husbandry and behavioral research. Despite significant progress in confined environments, applying these accomplishments in wild settings remains challenging. Thus, we create the first large-scale cattle face recognition dataset, ICRWE, for wild environments. It encompasses 483 cattle and 9,816 high-resolution image samples. Each sample undergoes annotation for face features, light conditions, and face orientation. Furthermore, we introduce a novel parallel attention network, PANet. Comprising several cascaded Transformer modules, each module incorporates two parallel Position Attention Modules (PAM) and Feature Mapping Modules (FMM). PAM focuses on local and global features at each image position through parallel channel attention, and FMM captures intricate feature patterns through non-linear mappings. Experimental results indicate that PANet achieves a recognition accuracy of 88.03% on the ICRWE dataset, establishing itself as the current state-of-the-art approach. The source code is available in the supplementary materials.

A Parallel Attention Network for Cattle Face Recognition

TL;DR

This work tackles cattle face recognition in wild environments by introducing the ICRWE dataset and a novel PANet architecture. PANet employs a parallel Transformer backbone with two attention pathways, PAM and FMM, to capture local and global features and generate discriminative 128-dimensional embeddings via triplet learning. On ICRWE, PANet achieves a state-of-the-art accuracy of , outperforming multiple SOTA methods while maintaining efficient model size. The dataset and code availability promise practical impact for wildlife monitoring and precision livestock management in uncontrolled environments.

Abstract

Cattle face recognition holds paramount significance in domains such as animal husbandry and behavioral research. Despite significant progress in confined environments, applying these accomplishments in wild settings remains challenging. Thus, we create the first large-scale cattle face recognition dataset, ICRWE, for wild environments. It encompasses 483 cattle and 9,816 high-resolution image samples. Each sample undergoes annotation for face features, light conditions, and face orientation. Furthermore, we introduce a novel parallel attention network, PANet. Comprising several cascaded Transformer modules, each module incorporates two parallel Position Attention Modules (PAM) and Feature Mapping Modules (FMM). PAM focuses on local and global features at each image position through parallel channel attention, and FMM captures intricate feature patterns through non-linear mappings. Experimental results indicate that PANet achieves a recognition accuracy of 88.03% on the ICRWE dataset, establishing itself as the current state-of-the-art approach. The source code is available in the supplementary materials.
Paper Structure (15 sections, 6 equations, 6 figures, 3 tables)

This paper contains 15 sections, 6 equations, 6 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Cattle Recognition
  4. Attention Mechanism
  5. Dataset
  6. Dataset Creation
  7. Dataset Properties
  8. Method
  9. Overall Pipeline
  10. Proposed Network
  11. Experiments and Results
  12. Comparison with SOTA Methods
  13. Ablation and Analysis
  14. Conclusion
  15. Acknowledgements

Figures (6)

  • Figure 1: Samples of cattle faces from the dataset exhibiting variations in lighting conditions and different face orientations.
  • Figure 2: Statistical results for four levels of lighting condition and three face orientations in the training and testing datasets.
  • Figure 3: The distribution and statistics of sample quantities for each cattle in both the training and testing sets.
  • Figure 4: The network framework employed in the method of cattle recognition based on cattle face features.
  • Figure 5: (a) Sequential Transformer structure; (b) Parallel Transformer structure; (c) The specific architectural implementation of the Position Attention Modules (PAM); (d) The specific architectural implementation of the Feature Mapping Modules (FMM).
  • ...and 1 more figures