We automate the monitoring of a pig’s health through images. We first apply a Region-Convolutional Neural Network (R-CNN) to segment the pigs in an image. Thereafter, we utilize a 12-layer Convolutional Neural Network (CNN) to classify cropped pictures either as that of a healthy pig or an unhealthy pig.
All this is done using a total of 236 images (129 healthy, 107 unhealthy) collected from an agricultural industry player and various online sources. Using a 75:25 train-test split ratio, we achieve accuracies of 94% and 92% on our train and test sets, respectively. Given the importance of the swine industry in the Philippines, we believe our work can be of great value to the agricultural industry and beyond.