Mao-Hsiang HuangEN-CHUNG LINYAN-FU KUO2019-11-042019-11-042019-01-01https://scholars.lib.ntu.edu.tw/handle/123456789/430860© 2019 ASABE Annual International Meeting. All rights reserved. The body condition of a sow indicates its degree of obesity and, hence, critically determines the health and productivity of the sow during the next pregnancy. Precisely scoring the body conditions of sows is essential in feeding management. Conventionally, the body conditions of sows are scored from rear views by breeders. Manual observation, however, largely relies on the experience of the breeders and can be subjective. This study aimed to score the body conditions of sows using image processing and deep learning. A convolutional neural network was developed to identify the bodies of sows in images. The aspect and conformation ratios of the sows were then determined. Body conditions were then scored based on these ratios. The study proved that image-based scoring of sow body conditions was achievable and could be applied to the livestock industry.Fully convolutional neural network | Image processing | Semantic segmentation | Sow body condition scoresFully convolutional neural network; Image processing; Semantic segmentation; Sow body condition scores[SDGs]SDG3Agriculture; Convolution; Deep learning; Image processing; Neural networks; Semantics; Body condition; Body condition score; Conformation ratio; Convolutional neural network; Image-based; Semantic segmentation; Image segmentationconference paper10.13031/aim.2019009152-s2.0-85072930448https://api.elsevier.com/content/abstract/scopus_id/85072930448