Labeling of an intra-class variation object in deep learning classification

Abstract

Machine orientation learning had demonstrated that deep learning (DL)-convolutional neural networks (CNNs) were robust image classifiers with significant accuracy. Although to been functional, DL scope classification as tight, well-defined as possible uses a 2-class object, for instance, cats and dogs. The DL classification faced many challenges, e.g., variation factors, the intra-class variation. This nature is presented in every object, its diversity of an object. The label was an exact given name of an intra-class variation object. Unfortunately, not every object had a specific name, in exceptionally high similarity inside the category. This paper explored those problems in flower plants’ taxonomy naming. In supervised learned of DL, image datasets musted labeled with a meaningful word or phrase that humans are familiar with, a taxonomy naming. Labeled with visual feature extraction brought a fully automatic classification. Flower Plumeria L labeling extracted from perspective dimension scale of petal flower which automatically obtained by contour detection, and peaks of blue green red (BGR) histogram channels from bins histogram after object masked. Dataset collected on photography workbench equipped with webcam and ring light. Results showed labels for intra-class variation of Plumeria L in form of dimension-scale and BGR-peaks. The result of this study presented a novelty in building datasets for intra-class variation for the DL classification.