Chahat Deep Singh

Projects

Segmenting Objects in Reconstructed Point Cloud from a Concave Structure multiple to Convex Structures.

Spring 2017

Iterative closest point (ICP) is a popular algorithm employed to register two sets of curves, two sets of surfaces, or two point clouds. The Iterative Closest Point (ICP) algorithm that uses the point-toplane error metric has been shown to converge much faster than one that uses the point-to-point error metric.

Understanding Geometry of Objects: Shape from Motion and Shading

Fall 2016

This project aim towards getting the shape, illumination, reflectance and other parameters from an image or a stream of images using Tomasi-Kanade’s and Barron-Malik’s method.

Reconnaissance Advanced Vehicle for Hostile Environment

Spring 2015

[]

The RAVHEN prototype is capable of leaping obstacles over 30cm which makes it easy to jump its way through the stairs.

REDIPS: A Quadruped Robot

Fall 2014

[]

Developed a Quadruped Robot, codenamed REDIPS having both autonomous and manual capabilities which followed Trot and Creep GAIT algorithm with human detection and object tracking.

Dynamic Rapidly-exploring Random Tree

March 2017

A rapidly exploring random tree algorithm designed to efficiently search high dimensional space to find an optimal path for a mobile robot.

Face Swapping

March 2017

Implemented Snapchat like face swap filter by performing face warping on the two images using Triangulation and Thing Plate Spline. The result was further processed using Poisson Blending and Motion Filtering to reduce the amount of flickering in the video.

My Camera Calibrator

February 2017

Finding a good initial estimate of the parameters i.e. solving for K matrix (intrinsics) and [R,t] (extrinsics) so as to feed it to a non-linear optimizer (Minimum Least Square).

Structure from Motion

April 2017

<Edit Description> Finding a good initial estimate of the parameters i.e. solving for K matrix (intrinsics) and [R,t] (extrinsics) so as to feed it to a non-linear optimizer (Minimum Least Square).

Seamless Carving

April 2017

Visual Flow in Dynamic Vision Sensor (DVS)

May 2017

This project introduces a new methodology to compute visual flow using an asynchronous event-based sensor.

Classifying CIFAR-10: Neural Networks vs Support Vector Machines

Fall 2016

This project aim towards the CIFAR-10 image classification using Support Vector Machines (SVM) using Local Binary Patterns (LBP) and Convolutional Neural Networks (CNN).

Collision-free Optimal Trajectory Planning in a Multiple Robot System

Fall 2016

Solving the problem of cooperative planning in order to simulate the algorithm for concurrent assignment and planning of trajectories (CAPT) for a team of robots.

Relevant Coursework

Image Understanding by Prof. Rama Chellappa (CMSC828P: Fall 2016)
Computer Processing of Pictorial Information by Prof. Yiannis Aloimonos (CMSC733: Spring 2017)
Machine Learning by Prof. Jordan Boyd-Graber (CMSC726: Fall 2017)
Perception for Autonomous Robots (ENPM673: Spring 2017)
Vision, Planning and Control in Aerial Robotics (CMSC828T: Spring 2017)
Planning for Autonmous Robots (ENPM661: Spring 2017)
Independence Study Research on Visual-Inertial Odometry for indoor navigation