Generating high-fidelity line drawings from cultural heritage artifacts is essential for archaeological and art historical documentation and analysis. This task is particularly challenging for ancient Egyptian artifacts due to limited training data and substantial stylistic diversity, ranging from densely patterned motifs to sparse hieroglyphs. We present Line-Density Aware Augmentation (LDAA), a novel data augmentation strategy that mitigates these challenges by analyzing the line density of ground-truth drawings and probabilistically synthesizing diverse, challenging training samples through the strategic recombination of high- and low-density regions from different images. Integrated with established generative frameworks like Pix2Pix and CycleGAN, LDAA enables models to effectively adapt to varying artistic styles and smooth transitions between them. Extensive qualitative and quantitative evaluations show that our approach significantly outperforms traditional methods, such as Canny edge detection, and baseline deep learning models, producing line drawings with higher structural accuracy and perceptual quality—even on severely eroded surfaces. Beyond providing a robust tool for cultural heritage preservation, LDAA offers a transferable augmentation paradigm applicable to other data-scarce computer vision domains.