Sanjaya Joseph Primary Author mixed material bibliography Bandung Maranatha University Press 2020 Indonesia

Artikel Jurnal

hlm : 311-323 Abstract 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. Informatika Sistem Informasi JUTISI 24432210 Perpustakaan Teknik UPI YAI JUTISI V6N2 Agustus 2020 JUTISI2-015 Perpustakaan FT UPI YAI JUTISI V6N2 Agustus 2020 cover_issue_174_en_US.png.png 59210 2023-02-27 14:04:00 2023-02-27 14:04:00 machine generated