Annotated video image dataset and analysis of suckling behaviors in newborn piglets

• Grant number: 24/19060-8
• Support Opportunities: Scholarships abroad - Research Internship - Post-doctor
• Start date: April 07, 2025
• End date: April 06, 2026
• Field of knowledge: Agronomical Sciences - Animal Husbandry - Animal Production
• Principal Investigator: Rafael Vieira de Sousa
• Grantee: Diego Feitosa Leal
• Supervisor: Tami Brown-Brandl
• Host Institution: Faculdade de Zootecnia e Engenharia de Alimentos (FZEA). Universidade de São Paulo (USP). Pirassununga , SP, Brazil
• Institution abroad: University of Nebraska-Lincoln (UNL), United States
• Associated to the scholarship: 23/10750-9 - Study on the influence of pre-partum activities pattern of sows on piglet survival using deep learning and computer vision, BP.PD

Abstract

High mortality rate of suckling piglets continues to represent a major source of economic loss and decreased welfare in the pig industry. Piglet mortality is a multifactor problem which is influenced by maternal, piglet and environmental factors. Low vitality at birth is a major factor contributing to pre-weaning mortality as it is associated with compromised colostrum intake. Piglet suckling behaviors can reflect vitality at birth. Currently, assessing newborn piglet suckling behaviors relies on human observation, which is time-consuming and prone to errors. Automated detection and analysis of suckling behaviors could be used as a strategy to early identify low-vitality piglets, reducing pre-weaning mortality. This BEPE project aligns with objective #3 of the postdoctoral scholarship in Brazil (grant number: 23/10750-9). The aims of this project are threefold: 1) to use an existing video dataset to annotate the suckling behaviors of individual piglets; 2) to evaluate the effect of three different farrowing crate layouts on newborn piglet suckling behaviors and performance in early lactation; and 3) to generate a labeled-image dataset for developing an automated monitoring system to predict piglet vitality and promptly identify piglets at risk for pre-weaning mortality. An existing video databank consisting of 546 farrowing processes will be utilized. The sows and piglets were housed in three different crate layouts during farrowing and lactation at the USDA-ARS Meat Animal Research Center, Clay Center, Nebraska, USA. The video recordings will be assessed, sorted, and edited. The script of an existing behavior analysis software will be modified to evaluate the following variables: interval from birth to first suckling, udder stimulation reflex, teat sampling, and udder post-massage. Selected video recordings will be uploaded to the software, and the above-mentioned variables will be annotated. The time to first suckle, behavioral time budgets, and piglet performance data will be organized in a spreadsheet and statistically analyzed. Creating a labeled-image dataset of piglet suckling behaviors will be the first step toward developing an automated computer vision monitoring system to predict piglet vitality and risk for preweaning mortality.