Auto Animal Detection and Classification among (Fish, Reptiles and Amphibians Categories) Using Deep Learning

Elham Mohammed Thabit A. ALSAADI and Nidhal K. El Abbadi

Neural networks (NN) have developed dramatically successful to solve prediction problems like, object recognition, scene understanding, and semantic segmentation which often surpassing accuracy of human level. This success is due to the ability of Deep Convolutional Neural Networks (DCNNs) to learn helpful representations of high dimensional inputs. In this research, a deep learning scheme based on convolutional neural network (CNN) has developed to detect and classify the object (animal) through training the system on images of three animal categories, like: Fish, Reptiles and Amphibians. We developed an algorithm which automatically extract features from images and use them in training process. The dataset used were consists of (6000 different image), where 4800 images used for training and 1200 images used for testing. From experiment we found that the best input image size is 40x40 and the best number of epochs is 100. The results are very good, and the accuracy of testing images reached to 99%. The results show that the proposed model has a positive effect on total performance of animal classification.

Volume 11 | 10-Special Issue

Pages: 726-736

DOI: 10.5373/JARDCS/V11SP10/20192863