Related Experiment Video
Updated: Jan 15, 2026
05:57
Long-term Video Tracking of Cohoused Aquatic Animals: A Case Study of the Daily Locomotor Activity of the Norway Lobster Nephrops norvegicus
Published on: April 8, 2019
7.2K
CF-DETR: A Lightweight Real-Time Model for Chicken Face Detection in High-Density Poultry Farming.
Bin Gao1,2, Wanchao Zhang2, Deqi Hao2
1Key Laboratory of Smart Breeding (Co-Construction by Ministry and Province), Ministry of Agriculture and Rural Affairs, Tianjin 300384, China.
Animals : an Open Access Journal From MDPI
|October 16, 2025
Summary
CF-DETR offers efficient chicken face detection for smart farming, improving accuracy in crowded conditions. This lightweight model enhances real-time monitoring in poultry production systems.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Agricultural Technology
Background:
- Automated monitoring in poultry systems requires reliable individual detection, especially in dense and cluttered environments.
- Existing methods may struggle with occlusion and background clutter, impacting efficiency in production-like settings.
Purpose of the Study:
- To develop an end-to-end object detector, CF-DETR, specifically for chicken face detection in challenging poultry production environments.
- To improve detection accuracy, reduce computational cost, and enhance robustness to occlusion and clutter.
Main Methods:
- CF-DETR builds upon the RT-DETR architecture, incorporating novel modules: Dynamic Inception Depthwise Convolution (DIDC), Polar Embedded Multi-Scale Encoder (PEMD), and Matchability Aware Loss (MAL).
- DIDC enhances receptive fields efficiently, PEMD restores global context and fuses multi-scale information, while MAL aligns confidence with localization quality.
Main Results:
- CF-DETR achieved high performance on a broiler dataset with mAP@0.50 of 96.9% and mAP@0.50-0.95 of 62.8%.
- The model demonstrated significant reductions in trainable parameters (33.2%) and FLOPs (23.0%) compared to the RT-DETR baseline, while maintaining a high speed of 81.4 FPS.
- Ablation studies validated the contribution of each module to performance and robustness.
Conclusions:
- CF-DETR provides a superior trade-off between detection performance and computational cost for real-time visual monitoring in intensive poultry production.
- Its lightweight design makes it suitable for deployment in real-time smart farming applications, addressing the need for efficient individual chicken detection.
