Identifying West African taurine cattle breed has become vital since uncontrolled crossing with zebus is jeopardizing their genetic heritage and trypanoresistance capacity. In this study, a computer vision solution is proposed for lobi taurine cattle classification. We implemented a customized Convolutional Neural Network (CNN) on a dataset containing 2379 images taken from three angles: front, side, and rear. The CNN was trained on four subsets of the image data according to the angle of shooting. The training is accelerated by a GeForce RTX3080 laptop GPU. The model yields only 78% precision for the mixed image dataset. Precision rises when the images are split by angle of view: 91% for side images and up to 99% for rear view images. Transfer learning has also been applied for comparison between our model and pretrained models. VGG-16 improved the results for all subsets.