Interactive Generation of 3D Volumetric Models from Point Clouds
Abstract - We present a volumetric modeling approach to interactively fit 3D primitives to point clouds in order to generate a CAD like model.Our model consists of two categories of volumetric primitives: n-gonal prisms and surfaces of revolution with B-spline profiles. By directly fitting volume primitives, our approach can effectively handle a great degree of noise. Our interactive approach compares favorably both with manually and automatically generated models.Not only is it much more time efficient than manual modeling but also gives significantly better output than state-of-the-art automatic method. Since the focal technique of our approach is the generation of 3D models at a primitive level, our results are ideal in the domain of architecture and preservation of heritage.
Mapping Rural Pakistan (MARUP)
This project was in collaboration with Technical University of Kaiserslautern, Germany. Its objective was automatic digitization of rural district maps of Pakistan. Digitization was a research challenge due to the limitations imposed by hand-drawn digits and degradation of maps. My contribution in this project was segmentation and recognition of Arabic numerals from scanned and stitched images of maps. By using image processing techniques, I extracted numerals and created feature vectors based on shape and context feature descriptors in Python. Experiments were performed using Neural Networks and Support Vector Machines (SVM) using WEKA.
Poisson Image Editing
This project was an implementation of Aseem Agarwala's "Efficient Gradient Domain Compositing using Quad-trees". It was a space and time efficient optimization for Poisson blending algorithm by using quad-tree data structure. By implementing quad-trees for subdivision of pixels, Poisson blending equations only had to be solved for pixels along the seams, which greatly improved the performance of the blending algorithm.
Implemented surface reconstruction of a fluid particle system using the method of Smooth Particle Hydrodynamics (SPH) and symmetric kernels for Computer Graphics class project.