Cattle Instance Segmentation by Transfer Learning Approach Using Deep Learning Models for Sustainable Livestock Farming

Citations Over TimeTop 10% of 2025 papers

Abstract

Image segmentation of animal instances is among the applications of artificial intelligence that recently emerged for sustainable livestock farming. This approach to sustainable livestock farming has become a practical avenue which several precision livestock farming researchers are utilizing to achieve their aims. Artificial intelligence involves several complex tasks including video and image processing. Moreover, the large volume of data required to train deep learning models makes these tasks more complex. However, with the emerging technology of transfer learning, these challenges have been remarkably mitigated. In this study, a system suitable for sustainable livestock farming is proposed using techniques of transfer learning and deep learning models for animal detection and recognition in a farm environment. We evaluated the proposed transfer learning method of the Enhanced Mask R-Convolution Neural Networks (CNN) by comparing it to other two deep learning models, namely Mask R-CNN and Faster R-CNN, on the same dataset using evaluation metric. The Enhanced Mask R-CNN obtained promising results of 0.2 s (computing time) and 97% (mAP), higher than the results obtained by the other two deep learning models. The findings in this study reveal the capacity of transfer learning as an approach for addressing the challenges in segmenting cattle images needed to improve precision livestock farming for sustainable agriculture.

Related Papers

• → Spatial distribution of livestock in Pakistan(1977)1 cited
• → Women, common property resources and livestock husbandry in Thai villages(1999)3 cited
• A Study on the Real-Time Livestock Monitoring System using Mobile Platform(2013)
• → Human‐Livestock Interaction(2004)47 cited
• Supply responses in the livestock economy of Kerala(1975)