All projects

StyleFusion — Neural Style Transfer

Hybrid neural style transfer pipeline combining classical optimisation with Stable Diffusion for higher-fidelity stylisation.

PythonPyTorchStable DiffusionHugging Face DiffusersVGG-19

Implemented a two-stage pipeline: VGG-based neural style transfer to capture coarse content/style alignment, followed by a Stable Diffusion img2img pass conditioned on style prompts to refine textures and coherence. Explored loss weighting, guidance scales, and denoising strengths to balance content preservation against style intensity on a curated set of artistic references.