Monitoring sheep species identification and classification
in the farming environment can be a tedious task and can be
a significant workload for a starting farmer. In this paper,
Convolutional Neural Network is proposed to reduce the
workload of sheep farmers. This experiment compares which
neural architecture model is more useful to classify sheep
species based on its face. The experiment was conducted using
the training dataset obtained from Kaggle. The dataset
contains 420 of each Marino sheep, Suffolk sheep, White
Suffolk sheep, and Poll Dorset sheep, totaling 1680 sheep face
images. This experiment was run on Google Colab, using the
Resnet50 network architecture model and VGG16 network
architecture model. The experiment shows good accuracy
results on the dataset achieving 86% using the Resnet50
network architecture model. Better accuracy results were
achieved using VGG16 network architecture, with an
accuracy value of 94%.