Pig3DNet: A 3D Convolutional Neural Network Model for Automated Pig Behavior Recognition Using Computer Vision

This study proposes an automated pig behavior recognition model based on 3D convolutional neural networks called Pig3DNet, which utilizes computer vision technology to achieve accurate recognition and analysis of pig behaviors. By constructing a multi-layer 3D convolutional structure, the model can effectively extract spatio-temporal features and recognize various pig behavior patterns including feeding, drinking, resting, moving, and others. Experimental results demonstrate that Pig3DNet achieves a behavior recognition accuracy of $\mathbf{9 4. 2 7 6 \%}$, recall rate of $\mathbf{9 3. 8 4 5 \%}$, and $\mathbf{F} 1$-score of $\mathbf{9 4. 0 6 1 \%}$, significantly outperforming traditional 2D convolutional neural networks and other machine learning methods. This model has important application value in intelligent livestock management, animal welfare monitoring, and precision farming.