Breeding Pig’s Butt Detection Based on SSD

Abstract

The experts can analyze much useful information from the breeding pig’s buttock. Nowadays, with the rapid development of information technology and computer vision, image feature engineering can also extract the features of the pig’s buttocks and analyze it. And butt detection is the first key to the intelligent analyze in the livestock environment. This study is focused on effectively identify each breeding pig’s buttock in the herd based on Single Shot Mutil-Box Detection (SSD). SSD is a convolutional neural network, which can easily trade-off the accuracy and time by one-stage detecting. This paper demonstrates the performance of SSD in livestock environment via many experiments. The mean average precision (mAP) of our algorithm reaches 77.04%, and it reaches 16 frames per second. In addition, all the dataset in the experiments of this study was collected at the Nutrition Institute of Sichuan Agricultural University. Therefore, the system can be easy to use in real-world applications.