Behavior recognition of cage-free multi-broilers based on spatiotemporal feature learning.

ABSTRACT: Poultry behavior indicates their health, welfare, and production performance. Timely access to broilers' behavioral information can improve their welfare and reduce disease spread. Most behaviors require a period of observation before they can be accurately judged. However, the existing approaches for multi-object behavior recognition were mostly developed based on a single-frame image and ignored the temporal features in videos, which led to misrecognition. This study proposed an end-to-end method for recognizing multiple simultaneous behavioral events of cage-free broilers in videos by Broiler Behavior Recognition System (BBRS) based on spatiotemporal feature learning. The BBRS consisted of 3 main components: the improved YOLOv8s detector, the Bytetrack tracker, and the 3D-ResNet50-TSAM

SUBMITTER: Hu Y 

PROVIDER: S-EPMC11470484 | biostudies-literature | 2024 Sep

REPOSITORIES: biostudies-literature