This README.md file was generated on 2017/01/16/ by Girum G Demisse.

This is a MATLAB implementation of a curved shape representation approach presented in:

• Demisse, G.G, Aouada,D, Ottersten, B. Similarity Metric For Curved Shapes in Eculidean Space., IEEE CVPR 2016.
• Demisse, G.G, Aouada,D, Ottersten, B. Deformation Based Curved Shape Representation., IEEE TPAMI 2017.
  

There is only one requirment for using the software: If you use this software (Curved_shape_representation) in its entirety or partially, please consider citing the papers listed above.

The software package is organized in three main folders:

• Classes - includes static and value classes.
• 3rdparty - includes codes developed by a third party.
• DP_opt - includes dynamic programming based optimizations.
• Dataset - includes Kimia's datasets for testing.

Curved shapes, extracted from Kimia99 and Kimia216, are included with the software package.

• Each column of "KIMIA99.mat" has 11 elements from the same shape category. In total, there are 9 shape catagories (columns).
• Each column of "KIMIA216.mat" has 12 elements from the same shape category. In total, there are 18 shape catagories (columns).

Three commented demo scripts are included with the package.

• Demo_HOWTO_curve_representation.m (I recommand you start from here)- shows how to:
  • Create a curve object-- build representation of a curve.
  • Estimate point correspondance, using either unfiorm or optimal sampling, between two curved shapes.
  • Compute geodesic curve and geodesic distance between two curved shapes.
  • Plot results.
• Demo_HOWTO_curve_model.m - shows how to:
  • Compute Karcher mean of a set of curved shape representations.
  • Compute K-clusters from a dataset of curved shape representations.
• Demo_HOWTO_deformation_transfer.m - shows how to:
  • Extract deformation that acts from the left, given two curved shapes.
  • Transfer a deformation to a given curved shape.

Girum G. Demisse, [email protected]