Abstract :
The diversity of reptiles and their uniqueness has increased the interest of
reptile enthusiast in various parts of the world, including Indonesia. This interest
does not only come from among the community of the enthusiast, but also from the
society due to the lack of knowledge of reptile species, which are too diverse in
shape and type, so that it creates interest for most people to know this fauna further
in order to increase knowledge. As the digital era develops, it takes artificial
intelligence for computer programs to do a human-like job where reptile species
can be identified automatically whenever a reptile image is given as input. Machine
learning through Deep Learning, especially the Convolutional Neural Network
(CNN) method, is needed for computer programs to identify the reptile species that
you want to know. This study aims to find the right model to produce high accuracy
in the identification of reptile species through input from images obtained manually
using a cellphone camera. The image is an image consisting of 3 RGB color
channels (Red, Green, Blue). Thousands of images are generated through the Data
Augmentation process of the collected images, resulting in tens of thousands of
training images. 8 models were tested in this study using the Python programming
language. With the use of the Dropout technique which is quite effective, the
prediction accuracy of 93% is obtained by this study as the highest accuracy in
identifying 14 different reptile species such as crocodiles, lizards, turtles and
crocodiles which are divided into the Crocodilia, Squamata and Testudinata
Orders.
Keywords: Data Augmentation, Species Identification, Convolutional Neural
Networks, Python, Reptile