Ankith Varun J  supervised by Prof. Anoop Namboodiri received his Master of Science – Dual Degree  in Computer Science and Engineering (CSD). Here’s a summary of his research work on Efficient Identity-preserving Face Swapping in Scenic Sketches:

We present an efficient framework for identity-preserving face swapping into scenic portrait sketches. First, we analyze latent identity representations via a StyleGAN-based cartoonization pipeline and show that identity encodings shift significantly across domains, motivating the need for domain-aware modelling. Second, we introduce Portrait Sketching StyleGAN (PSStyleGAN), which uses attentive affine-transform blocks to adapt a pretrained StyleGAN’s style codes, enabling rapid, few-shot learning of new sketch styles while faithfully preserving identity and editability. Third, we extend diffusion-based generative modelling by adapting InstantID for sketch synthesis, integrating strong semantic identity embeddings and landmark guidance to achieve zero-shot, high-fidelity personalized sketches. Finally, we build a real-time, endto-end pipeline for tourist applications, combining optimized sampling schedules and pipelinelevel StreamDiffusion techniques to deliver personalized scenic sketches in under five seconds. Extensive evaluations confirm that our methods substantially improve identity preservation and visual quality over prior approaches.

  November 2025