International Journal on Advanced Science, Engineering and Information Technology, Vol. 10 (2020) No. 6, pages: 2290-2296, DOI:10.18517/ijaseit.10.6.10948

MobileNets: Efficient Convolutional Neural Network for Identification of Protected Birds

Yulius Harjoseputro, Ign. Pramana Yuda, Kefin Pudi Danukusumo


Wildlife trade is one of the main factors causing endangered bird species. In Indonesia, trade has caused 28 bird species to be classified in the endangered bird category. Protection efforts have been made with the establishment of 564 species of Indonesian birds as protected birds. For law enforcement, certainty is needed in the identification of these bird species. This study begins with a Forum of Discussion Groups from relevant institutions in Java and Bali to determine the types of protected birds that are prioritized to developed in this application. Based on the results of the Forum of Discussion Group, a bird photo dataset compiled using 17 categories or types of bird photos as prioritized in this study. The method used in this study is the Convolutional Neural Network (CNN) method, which combined the structure of MobileNet and the weight of the network that has previously trained using ImageNet. The results of this study are the differences of results between CNN standards and those combined with the structure of MobileNet. For better accuracy, using the CNN standard, which is around 98.38% for the accuracy of the training, while in terms of size, combined with MobileNet has a relatively smaller model size, which is 68 megabytes.


wildlife; endangered; forum of discussion group; dataset; convolutional neural network; MobileNet; ImageNet.

Viewed: 1388 times (since abstract online)

cite this paper     download