My project presents a novel pipeline for high-fidelity 3D asset creation from a single image, combining state-of-the-art methods in geometry reconstruction and material segmentation. The process begins with SF3D, which reconstructs a detailed 3D mesh and extracts an albedo map from the input image, enabling accurate surface geometry and base colour information. The output is then processed by MaterialSeg3D, which segments and classifies material regions to generate an ORM (Occlusion, Roughness, and Metallic) map suitable for physically based rendering. This sequential pipeline not only streamlines 3D asset creation but also enhances realism by disentangling illumination from intrinsic object properties and applying learned material priors to enrich texture and shading.