Augmentasi Data Pengenalan Citra Mobil Menggunakan Pendekatan Random Crop, Rotate, dan Mixup

Abstrak

Deep convolutional neural networks (CNNs) have achieved remarkable results in two-dimensional (2D) image detection tasks. However, their high expression ability risks overfitting. Consequently, data augmentation techniques have been proposed to prevent overfitting while enriching datasets. In this paper, a Deep Learning system for accurate car model detection is proposed using the ResNet-152 network with a fully convolutional architecture. It is demonstrated that significant generalization gains in the learning process are attained by randomly generating augmented training data using several geometric transformations and pixel-wise changes, such as image cropping and image rotation. We evaluated data augmentation techniques by comparison with competitive data augmentation techniques such as mixup. Data augmented ResNet models achieve better results for accuracy metrics than baseline ResNet models with accuracy 82.6714% on Stanford Cars Dataset.