An End-to-End Model for Chicken Detection in a Cluttered Environment

Animals in the livestock business need to be seen in real time and online in order for them to grow and develop. The growing global demand for meat and poultry, the development of the livestock and poultry business, and the expansion of their region have made it more difficult for humans to handle the task due to human mistake and the growing amount of work. But thanks to technological breakthroughs like artificial intelligence, learning models, and closed-circuit cameras, most tasks may now be left to robots, preventing human error. This system may be installed and connected worldwide, allowing it to count all animals at any one time and give us information on each animal’s number and location.In this research, we created a model consisting of a segmentation module that readies input data for a detector model based on YOLOv8. In order to study chicken behavior, an integrated end-to-end approach is utilized to track and count the birds. The experimental results provide insight into the new proposal’s accuracy performance compared to earlier methods, demonstrating an 8% improvement. The goal of this article is to teach a YOLOv8-based learning model that may be applied to the development of an automated, integrated system for counting and assessing animal welfare in the future.