Aggressive Behavior Recognition for Group-Housed Pigs

EasyChair Preprint 15928

Abstract

This paper proposes a method for automatically identifying aggressive behavior in pigs using fixed-position monitoring, featuring a hybrid architecture that inte-grates object detection, tracking, and behavioral analysis. In the proposed archi-tecture, a stationary camera captures images, with YOLO detecting pig locations and DeepSORT tracking their movements to identify individuals potentially ex-hibiting aggression. This process generates five-second video clips of individual pigs, which are then processed by a behavioral analysis module based on Convo-lutional Neural Network (CNN) and Long Short-Term Memory (LSTM) net-work. Experimental results demonstrate that the proposed method achieves ap-proximately 90% accuracy in recognizing aggressive behavior in pigs on the test dataset.

Keyphrases: DeepSORT, LSTM, Penned pigs, YOLO, behavior recognition, deep learning

@booklet{EasyChair:15928,
  author    = {Chia Duo Wang and Yea Shuan Huang and Chang Wu Yu},
  title     = {Aggressive Behavior Recognition for Group-Housed Pigs},
  howpublished = {EasyChair Preprint 15928},
  year      = {EasyChair, 2025}}