Identifikasi Jenis Anjing Berdasarkan Gambar Menggunakan Convolutional Neural Network Berbasis Android

Kevin Oktovio Lauw (Program Studi Informatika)
Leo Willyanto Santoso (Program Studi Informatika)
Rolly Intan (Program Studi Informatika)

Jurnal Infra
Vol 8, No 2 (2020)

Dogs are raised by many people however, to maintain a dog, there are several factors that must be considered such as feed consumed, intensity of care, and cleanliness of the cage or the appropriate environment. Therefore, an android application is needed to identify the type of dog and provide information related to the type of dog. The method we used is You Only Look Once to detect dog objects in an image then the dog image is cropped, the results will be processed by the Convolutional Neural Network to identify the type of dog based on the image given after that displaying the results of its identification on android. The test results show that the identification results from CNN are very dependent on the results of predictions from YOLO because the input from CNN is the result of predictions from YOLO. YOLO has a disadvantage where it will detect dog dolls and dog fur as dog objects. The test results show the accuracy of YOLO to detect dogs is 94.242%, CNN accuracy of model I is 56.400%, accuracy of CNN model II is 40,000% and accuracy of CNN model III is 50.400%.