Projects Overview:
User Hand/Finger Tracking using Kinect: This project is part of my Directed Research with the Medical VR Group at the Institute for Creative Technologies (led by Dr. Belinda Lange). This project focuses on using the Microsoft Kinect SDK to track hand opening/closing and finger movements and includes a mini game in Unity3D Game Engine to demonstrate hand tracking movements as a Rehabilitation Game.Check the progress here : http://vimeo.com/album/1884887
3D Face Modeling and Recognition: Worked on the research project to develop middleware for OpenNI for 3D Face Recognition under the advisement of Dr Jongmoo Choi at USC Institute for Robotics and Intelligent Systems. This uses 3D face data and depth information for detection and recognition of the face. Uses robust methods like ICP for registration of probe face with reference and gallery. The software is free for download from OpenNI site here: http://openni.ru/files/3d-face-identification/index.html
Multi User tracking using multiple Kinects: This is an ongoing research project to track people using multiple Microsoft kinects using OpenNI API on ROS under the guidance of PhD student Ross Mead. We are currently building error models between kinect sensed data to recognize the same skeleton. The current progress available in video on vimeo, below are the links. http://vimeo.com/37363798 and http://vimeo.com/37363828.
Duplo Point Cloud Extraction : The aim of the project was to extract individual point clouds for the scenario where identical colored duplos are kept in contact with each other. The procedure involved segmenting the input cloud using planar segmentation, then pass-through filtration ,k-mean color based coloring to handle generic cases followed by Euclidean distance clustering to separate the duplos which are placed at a distance from each other (using radius search). To handle the complex cases of duplos placed one on each other Sample consensus Initial Alignment is used with a prior of knowledge of duplo’s height and width. This was implemented using PCL over ROS. The project report is here and presentation here.
Implementing Renderer from Scratch(Computer Graphics): I have implemented a Renderer from scratch using a language called Processing developed by MIT. The renderer is capable of processing asc files , raster it using LEE algorithm. The renderer is capable of doing Trasformations(Rotation,Translation,Scaling), Handles ambient,directional and specular lightning, Light source animation, Camera rotation, Flat Shading, Toon Shading, Phong Shading, Gouraud Shading,Perspective correction, Texture Mapping, Procedural Texturing like perlin, turbulence,Anti Aliasing.
Kinect Bridge: Capturing data from kinect on windows and publishing the messages over network for an ubuntu client.
Artificial Intelligence Implementations: NConnect Game, N-Queen Problem with user and human(using minimax and alphabeta), 8-puzzle game, Einstein problem using prolog.
3D Face Modeling and Recognition: Worked on the research project to develop middleware for OpenNI for 3D Face Recognition under the advisement of Dr Jongmoo Choi at USC Institute for Robotics and Intelligent Systems. This uses 3D face data and depth information for detection and recognition of the face. Uses robust methods like ICP for registration of probe face with reference and gallery. The software is free for download from OpenNI site here: http://openni.ru/files/3d-face-identification/index.html
Multi User tracking using multiple Kinects: This is an ongoing research project to track people using multiple Microsoft kinects using OpenNI API on ROS under the guidance of PhD student Ross Mead. We are currently building error models between kinect sensed data to recognize the same skeleton. The current progress available in video on vimeo, below are the links. http://vimeo.com/37363798 and http://vimeo.com/37363828.
Duplo Point Cloud Extraction : The aim of the project was to extract individual point clouds for the scenario where identical colored duplos are kept in contact with each other. The procedure involved segmenting the input cloud using planar segmentation, then pass-through filtration ,k-mean color based coloring to handle generic cases followed by Euclidean distance clustering to separate the duplos which are placed at a distance from each other (using radius search). To handle the complex cases of duplos placed one on each other Sample consensus Initial Alignment is used with a prior of knowledge of duplo’s height and width. This was implemented using PCL over ROS. The project report is here and presentation here.
Implementing Renderer from Scratch(Computer Graphics): I have implemented a Renderer from scratch using a language called Processing developed by MIT. The renderer is capable of processing asc files , raster it using LEE algorithm. The renderer is capable of doing Trasformations(Rotation,Translation,Scaling), Handles ambient,directional and specular lightning, Light source animation, Camera rotation, Flat Shading, Toon Shading, Phong Shading, Gouraud Shading,Perspective correction, Texture Mapping, Procedural Texturing like perlin, turbulence,Anti Aliasing.
Kinect Bridge: Capturing data from kinect on windows and publishing the messages over network for an ubuntu client.
Artificial Intelligence Implementations: NConnect Game, N-Queen Problem with user and human(using minimax and alphabeta), 8-puzzle game, Einstein problem using prolog.