Deteksi Ujung Jari menggunakan Faster-RCNN dengan Arsitektur Inception v2 pada Citra Derau

Abstract

Fingertip detection is a field on computers that has extensive space in field: NUI, robotics, etc. CNN is one method that is being used in object detection, with some CNN updates being faster - RCNN is able to detect objects very well. This study conducted the ability of Faster-RCNN in detecting fingertips with the Inception V2 architecture. Implementation is done on images that have noise and not. The results showed that image without noise has 91% accuracy, while for each noisy image: Gaussian, Salt and Pepper, Poisson and Speckle had an accuracy of 34%, 5%, 80% and 21%.