Performance Comparison of VGG16 and ResNet Architectures on CIFAR-10 Dataset

Abstract In this paper, we present a comparative analysis of three widely used convolutional neural network architectures: VGG16, ResNet50, and ResNet50V2. The models are trained and evaluated on the CIFAR-10 dataset using standard performance metrics such as accuracy, precision, recall, F1-score, and AUC. Experiments were conducted using Google Colab and TensorFlow/Keras frameworks. The results demonstrate that the VGG16 model consistently achieves higher classification accuracy and area under the curve (AUC), outperforming both ResNet variants. Specifically, VGG16 achieves an accuracy of 89% and an AUC of 0.9909 after 20 epochs. This study provides practical insights into architecture selection for small-scale image classification tasks.