Comparative Evaluation of CNN Architectures for Neural Style Transfer in Indonesian Batik Motif Generation: A Comprehensive Study

TL;DR

This study systematically compares five pretrained CNN backbones (VGG16, VGG19, Inception V3, ResNet50, ResNet101) for neural style transfer in Indonesian batik motif generation. Using 245 controlled experiments on constrained hardware, it shows no significant differences in structural preservation (ANOVA on SSIM $p=0.83$) across architectures, while ResNet variants dramatically improve training speed (about $5$–$6\times$ faster) and reduce FLOPs to $0.63$ GFLOPs from $10.12$ GFLOPs for VGG, without compromising perceptual similarity (LPIPS around $0.53$ for ResNet vs $0.63$–$0.65$ for VGG). Qualitative analyses reveal VGG yields denser painterly textures, whereas ResNet prioritizes structural fidelity and milder stylization, with Inception V3 offering an intermediate but noisier style. The findings shift NST architectural emphasis toward efficiency-aware deployment and structure-preserving stylization, recommending ResNet-based backbones as practical baselines for scalable batik motif generation, complemented by hyperparameter tuning to achieve desired aesthetics.,

Abstract

Neural Style Transfer (NST) provides a computational framework for the digital preservation and generative exploration of Indonesian batik motifs; however, existing approaches remain largely centered on VGG-based architectures whose strong stylistic expressiveness comes at the cost of high computational and memory demands, that limits practical deployment in resource-limited environments. This study presents a systematic comparative analysis of five widely used CNN backbones, namely VGG16, VGG19, Inception V3, ResNet50, and ResNet101, based on 245 controlled experiments combining quantitative metrics, qualitative assessment, and statistical analysis to examine the trade-off between structural preservation, stylistic behavior, and computational efficiency. The results show that backbone selection does not yield statistically significant differences in structural similarity, as confirmed by ANOVA on SSIM (p= 0.83), indicating comparable levels of structural preservation rather than equivalent stylistic quality. Within this context, ResNet-based architectures achieve approximately 5-6x faster convergence than VGG models while maintaining similar perceptual similarity (LPIPS = 0.53) and requiring over 16x fewer FLOPs (0.63 vs 10.12 GFLOPs). Qualitative analysis reveals consistent stylistic trade-offs, with VGG producing denser painterly textures, ResNet favoring geometric stability and canting stroke preservation with milder stylization, and Inception V3 exhibiting intermediate but noisier behavior. These findings reposition architectural choice in NST from maximizing stylistic intensity toward efficiency-aware and structure-preserving deployment, highlighting ResNet-based backbones as a practical foundation for scalable, industry-oriented batik generation.

Figures (9)

• Figure 1: Architecture of VGG-16
• Figure 2: Architecture of ResNet
• Figure 3: Architecture of Inception
• Figure 4: Box plot distribution of quantitative performance metrics (SSIM, PSNR, and LPIPS) across different CNN architectures.
• Figure 5: Mean Comparison Plot of quantitative performance metrics (SSIM, PSNR, and LPIPS) across different CNN architectures.