Automatic recognition of pig body parts using YOLO V5 for disease and abnormality monitoring
Md Nasim Reza (a, b), Md Sazzadul Kabir (a), Eliezel Habineza (a), Sumaiya Islam (a), Minho Song (c), Gookhwan Kim (d), Sun-Ok Chung (a,b*)

(a) Department of Smart Agricultural Systems, Graduate School, Chungnam National University, Daejeon 34134, Republic of Korea
(b) Department of Agricultural Machinery Engineering, Graduate School, Chungnam National University, Daejeon 34134, Republic of Korea
(c) Department of Animal Science and Biotechnology, Graduate School, Chungnam National University, Daejeon 34134, Republic of Korea
(d) National Institute of Agricultural Sciences, Rural Development Administration, Jeonju 54875, Republic of Korea

*Corresponding author: Sun-Ok Chung, E-mail: sochung[at]cnu.ac.kr

Abstract

Recent years have seen a rise in the significance of both the monitoring of farm animals and the automated detection of the health state of pigs, both of which play an essential role in the study of agricultural practice. This research was conducted with the intention of developing a method that could automatically distinguish various parts of pig^s bodies by making use of a two-dimensional image-based deep neural network (DNN) for the purpose of disease and abnormality monitoring in pigs farm. Yolo V5 algorithm for recognizing objects was used in the process of training the automated identification of pig^s body parts in the farm condition. In order to acquire images of the pigs, an automatic image acquisition system was established in the pig farm. From the obtained photos, 206 heterogeneous images were extracted and 12,123 bounding box annotations were made to form a high-quality dataset. The annotations of the bounding box were then utilized to train the proposed model. Our proposed algorithm achieved an average precision (AP) of 0.85% to 94% of different body parts of pigs. Overexposed or blurry images have poor detection accuracy. Results indicate that the approach suggested in this research recognizes pig body parts adequately in a pig farm setting and may be useful for livestock farm applications.

Keywords: Pig detection, pig body parts, diseases monitoring, computer vision, Yolo

Topic: Agricultural engineering