Swarna   Kamlam   Ravindran

Projects: Computer Vision

DST Project on Features for 3D applications (Dec 2012 - Dec 2014)

  • CoMaL: Corners on Maximally-stable Level line segments, Swarna Kamlam Ravindran and Anurag Mittal, IEEE Conference on Computer Vision and Pattern Recognition, 2016. [Bibtex] [PDF] [Code]

  • The performance of the feature detection and matching modules were significantly improved at the object boundary regions, useful in applications such as Vehicle Tracking and Structure from Motion.
  • Traditional gradient-based features fail at depth discontinuities due to object motion against a varying background. Our features work on the curve describing the object leading to more accurate tracking.
  • The project was funded by the Department of Science and Technology, Govt of India.

DRDO Project on Features for Surveillance (Dec 2009 - Dec 2010)

  • A feature detector using a probabilistic combination of stable extremal regions incorporated into a 3D reconstruction system yielded over 15% improvement over existing methods for surveillance.
  • Mosaicing and Contour detection software were implemented by using shape geometry to achieve scale invariance.
  • The project was for the Defense Research and Development Organisation.
  • Scale-invariant curve-based features for tracking under varying backgrounds, Swarna Kamlam Ravindran and Anurag Mittal, Tech Report, DRDO
  • CMSER: Combined MSERs for better feature matching, Swarna Kamlam Ravindran and Anurag Mittal, Tech Report, DRDO.

3D Face Recognition, (Dec 2008 - May 2009)

  • 3D Face Recognition system using a Local Shape Descriptor, Swarna Kamlam Ravindran and Sumithra G, International Symposium on Computing, Communication, and Control (ISCCC) 2009.
  • Designed a face recognition system using a descriptor formed as a 2D histogram of distances between points near a landmark point to the corresponding tangent plane. Experiments were performed on range images of six subjects with varying illumination.

Bag detection for TCS surveillance project (Jul 2008 - Dec 2008)

  • Principles of symmetricity were used to identify humans carrying bags or other objects. Periodocity check was employed to identify vehicles or animals in motion.

Modules for Music & Speech projects, IIT (Nov 2011 - Feb 2012)

  • Data classification was performed using Linde Buzo Algorithm and K means clustering.
  • Bayesian Clustering and Bayesian classification using BE, ML, MAP was implemented to classify high-dimensional data points. [Code]

Projects in Computer Vision (May 2011 - Jul 2011)

  • Codes for Image Mosaicing, Image Rectification and Parabola detection using Hough transform were developed. [Code]

Projects in Advanced DSP (May 2011 - Jul 2011)

  • Non-stationary analysis on audio signals was performed using spectrogram, scalogram and Wigner-Ville distributions. Tools to analyse an ECG signal and perform Signal Quantisation were developed. [Code]

Projects: Coding Theory

Coding in the DNA (Dec 2012 - Aug 2013)

  • Investigated the presence of an error correction code in the DNA of E.coli by modeling evolution of the DNA as a communication process and mutations as errors. [Code]

Threshold Decoder for TeNet project, IIT(Jan 2012 - Aug 2012)

  • Built a threshold decoder for Soft Linear Block Codes using Bossert Hergert Algorithm.
  • Built decoders for Hamming and Reed Muller codes. [Code]

Decoders for TeNet project, IIT-Madras (Jan 2012 - Aug 2012)

  • Built generic encoder-decoders for different types of LDPC codes and channels.
  • Built a generic Viterbi decoder for convolution codes.[Code]