Grad School Projects (at Guibas Lab, Stanford University)

My work in the Guibas Lab, has been focussed on trying to understand relationships between different shapes by means of shape correspondences. While this is a problem that is well studied in the realm of isometric shapes, nonisometric shapes are a different beast altogether. Studying shape correspondences for nonisometric shapes can be done with a strong assumption on the shape types. These assumption could be such as that of geometric similarity, or even grammatical nonvariability in the shape structure of these nonisometric shapes.

Generative Grammars for Shapes

The project on Shape Grammars is aimed at generating new shape models from pre-loaded shape databases utilizing existing information about the grammar of a certain shape model and being able to introduce a certain dimensional variability in this model. This consists of two stages: (1) Grammar induction from a given shape source, say a point cloud model (2) Learning dimensions of the existing point cloud models on a certain grammar. The applications of a setup that can perform these tasks are multifold such as shape completion, decomposition and understanding.

Stable Shape Correspondences between Non-Isometric Shapes

We consider the problem of finding meaningful correspondences between 3D models that are related but not necessarily very similar. When the shapes are quite different, a point-to-point map is not always appropriate, so our focus in this paper is a method to build a set of correspondences between shape regions or parts. The proposed approach exploits a variety of feature functions on the shapes and makes use of the key observation that points in matching parts have similar ranks in the sorting of the corresponding feature values. Our algorithm proceeds in two steps. We first build an affinity matrix between points on the two shapes, based on feature rank similarity over many feature functions. We then define a notion of stability of a pair of regions, with respect to this affinity matrix, obtained as a fixed point of a nonlinear operator. Our method yields a family of corresponding maximally stable regions between the two shapes that can be used to define shape parts. We observe that this is an instance of the biclustering problem and that it is related to solving a constrained maximal eigenvalue problem. We provide an algorithm to solve this problem that mimics the power method. We show the robustness of its output to noisy input features as well its convergence properties. The obtained part correspondences are shown to be almost perfect matches in the isometric case, and also semantically appropriate even in non-isometric cases. We provide numerous examples and applications of this technique, for example to sharpening correspondences in traditional shape matching algorithms. PDF

Undergraduate (Dual Degree) Projects

Distributed Storage For Quasi-Cyclic Codes

I worked on Finding efficient regenerative (n,k) quasi cyclic codes for distributed storage for my dual degree thesis at IIT Madras. This involves finding regenerative vectors for a (n,k) quasi cyclic code, which is used to encode across a distributed storage system with n server nodes. These vectors are needed to minimize the information (bits) transferred to restore a failed node. This involves the ideas of interference alignment makes use of Galois fields. This work was advised by Dr.Andrew Thangaraj of IIT Madras. The work was submitted to International Symposium in Information Theory (ISIT)- 2013, and was presented in Istanbul, Turkey. PDF.

MAX Function Computation-Optimal Timer Based Selection Algorithms

My work with Dr.Neelesh Mehta, at the Department of Electrical Communication Engineering, IISc Bangalore, involved max-function computation for a double tier regular tree, using optimal timer based selection algorithms. The idea was to identify the node, containing the max-function value amongst all the nodes, in a two-tier regular tree. The theory of optimal timer based selection algorithms was used. This was compared with the more obvious polling scheme, and was shown to be much better in performance.

GPS/GLONASS Simulator

The project I undertook as a summer intern at Texas Instruments India, was to develop a C++ simulating framework for their existing GPS/GLONASS setup. This involved understanding the various intricacies of how the GPS worked, including the acquisition and tracking phases. The information carried by the GPS had to be quantized, and this was the bottleneck of the process. I also worked on setting up the system base connecting the acquisition and tracking phases, and passing on the acquired information to the tracking phase.

Touch-Talk

My first project in IIT Madras was "Touch-Talk", an inter-disciplinary project, which was part of a spirit of Engineering (SoE) initiative by the Centre for Innovation (CFI), IIT Madras. This project,which was done in collaboration with Prasad Sarangapani, Shaileshh Bojja Venkatakrishnan, Vasuki Narasimha Swamy and Bharath Kumar was aimed at converting touch input on a touch screen to speech output for Indian languages. This project was guided by Dr.Srinivasa Chakravarthy of the Department of Biotechnology, IIT Madras, and we also were heavily helped by Dr.C.S.Ramalingam of the Department of Electrical Engineering, IIT Madras, and Dr.Hema Murthy, the Department of Computer Science and Engineering, IIT Madras, in the process. This project was heavily based on gesture recognition, and writing C++ code to convert recognised touch to speech.