Ai-Powered Pencil Strokes: Face Sketch Synthesis Using Cyclegan and Style Transfer

This paper introduces an AI-driven framework for face sketch synthesis using Cycle Generative Adversarial Networks (CycleGAN), enhanced with a style transfer module to improve the perceptual quality of generated sketches. Unlike conventional edge-detection or handcrafted filter-based techniques, our method leverages adversarial learning to capture intricate details, shading, and texture while maintaining structural coherence. By incorporating perceptual and edgeaware constraints into the objective function, the model enhances sketch quality. Qualitative analysis reveals progressive refinement over training epochs, while quantitative evaluation using the Structural Similarity Index (SSIM) confirms improved structural and perceptual similarity. The stable convergence of training loss further validates the model's effectiveness. Experimental results demonstrate that our approach generates highly realistic and visually appealing sketches, surpassing traditional methods in both fidelity and perceptual quality. Future work will focus on enhancing fine details and improving generalization across diverse datasets.