AVITRÓN: Selective Feeding Station for Free-Range Hens Based on YOLOv8 and Raspberry Pi

Intelligent automation in poultry production serves as a strategic pillar for enhancing efficiency, sustainability, and animal welfare in rural systems. In this study, AVITRÓN was developed as an autonomous station integrating computer vision and embedded artificial intelligence, designed for selective feed dispensing in free-range hens. The system combines a Raspberry Pi 5 with a 12 MP AI camera, an MG996R servomotor, and a YOLOv8-nano model trained on 402 images, expanded through data augmentation to 966 effective samples. The model achieved mAP@0.5 = 0.96, mAP@0.5:0.95 = 0.87, and F1 = 0.94 on the internal validation set, with an average latency of 175 ± 30 ms per frame (640 × 640 px), demonstrating its suitability for edge computing applications. During independent prototype validation, conducted with 144 external images excluded from training, the system operated continuously throughout the experimental test and completed 60 effective dispensing cycles. The system achieved an accuracy of 0.986, a precision of 1.000, and a recall of 0.968, maintaining stable performance under heterogeneous rural conditions. These results indicate that integrating lightweight artificial intelligence with embedded hardware represents a viable pathway toward the sustainable automation of rural poultry systems. AVITRÓN emerges as an accessible and scalable precision agriculture tool aligned with the Sustainable Development Goals (SDGs 2, 9, and 12), promoting more efficient and responsible production practices in rural environments.