A research team led by Guoming Li, PhD, assistant professor in the University of Georgia’s poultry science department, developed a user-friendly, web-based artificial intelligence (AI) system to analyze the broiler activity index (BAI).
Broiler activities like locomotion and movement can be quantified as the activity index, which is automatically computed by determining changes in broiler-representing pixels between adjacent images.
The BAI has been associated with broiler leg health, productivity and physical conditions. However, the index requires engineering knowledge to perform such tasks as adjusting image-processing parameters, segmenting individual broilers and selecting calculation areas. The current calculation procedures are not user-friendly for producers or animal scientists who do not have image-processing experience.
Study design
The project’s goal was to develop a web-based AI system for analyzing the BAI with three specific objectives:
- Verify the biological meaning of the activity index
- Explore efficient algorithms to segment individual birds from images
- Develop a user-friendly interface for analyzing BAI
The research team classified the BAI into high, medium and low using machine-learning models. They also used Streamlit, an online library for app development, to build a user-friendly platform to calculate the BAI, either individually or in groups, from videos.
Key takeaways
High- and medium-activity levels were significantly lower for broilers with cyclical heating operations than those without heating operations. In addition, the modified general deep-learning model, without extensive training, demonstrated an ability to achieve satisfactory performance (> 84% accuracy) in segmenting birds from poultry-housing images.
“The results indicated that the BAI can effectively indicate broiler heat stress, which can support early and timely interventions to improve broiler performance,” Li concluded.
He added that the open-source, user-friendly platform allows researchers to interact with software tools to understand animal behavior patterns and welfare without needing extensive programming knowledge. Also, Li explained, detailed quantitative measurements of poultry behaviors via the tool can better unravel treatment effects and help produce broilers with better well-being and productivity.
“Company technicians can access and deploy the tool to develop low-cost commercial products for automatic monitoring, further improving the automation level of the poultry industry and reducing labor for flock inspection,” Li said.
The research was funded by USPOULTRY and the USPOULTRY Foundation. Click here to view the industry summary.
Editor’s note: Content on Modern Poultry’s Industry Insights pages is provided and/or commissioned by our sponsors, who assume full responsibility for its accuracy and compliance.
