Single-Image 3D Reconstruction Pipeline
SRM Institute of Science and Technology, 2022
A cost-effective pipeline for 3D reconstruction from single 2D images, developed for SLAM and VR applications. I led its development in a three-member undergraduate research team advised by Dr. T. Rama Rao. The pipeline uses SIFT and ORB feature matching and epipolar geometry calculations, implemented and benchmarked with OpenCV, PyTorch, and Blender. Optimized projection matrices improved reconstruction accuracy beyond traditional optimization baselines, while device-agnostic image-processing techniques enabled rapid generation of 3D assets suitable for autonomous navigation systems.