The equine industry represents a very interesting economic activity worldwide, generating an estimated movement of 300 billion US dollars during 2017. Therefore, the scientific community has started developing different fields of investigation regarding data analytics to help improve growth in regard to this sector.
The study presents a preliminary analysis on the different public datasets regarding the equine industry as well as its taxonomy. The starting point of the study is the problem regarding classifying images for six horses breeds Atabay2017, by implementing Deep Learning (DL).
The staring point results provided by Atabay2017, showed a 95.9% accuracy on the test partition for the dataset. In the present research, Atabay results have been improved using a personalize convolutional DL model.
To obtain the current model Convolutional preexistent DL nets (Inception, Xception, REsnet, VGG16 y VGG19) have been compared with several animals incrementally entered, using transfer learning together with the dataset of the six breeds of horses used for Atabay 2017.
The results with this approach have improved the starting result from Atabay raising accuracy for up to 97.31%