Citra Sitentik Untuk Klasifikasi Buah Menggunakan Algoritma SIFT Descriptor, Bag of Features dan Support Vector Machine

Recognizing specific objects assigned to a computer using artificial intelligence of course goes through a training and testing process using machine learning methods, the limited number of datasets makes it difficult for deep learning methods to carry out classification, so to overcome this, other methods are needed, including Scale Invariant Features Transform ( SIFT) which is a method of image processing to extract features from a limited amount of data and combined with a method in machine learning. To overcome the inability of deep learning to use limited datasets, this research uses a combination of SIFT and bag of features to extract features and support vector machine (SVM) to carry out classification. In this study, the aim is to observe the effect of synthetic images on the performance of the combination of SIFT descriptor, Bag of Features and Support Vector Machine algorithms in classifying real fruit images. The dataset involved is a synthetic image in the form of a 3D image that is made into a complete object, then taking random views to make an image that represents the object as training data. Furthermore, for testing data, real images taken from the dataset link in previous research will be used. The number of synthetic datasets that can be collected for each fruit is 150 images, so that the total is 450 images, while the real fruit images consist of 148 apple images, 152 banana images, and 166 orange images, so that the total real images are 466 images. The results of this research show that the highest accuracy was 65.45% with an F1-score reaching 58.45%.