readme

Shape registration by shape context: algorithm implemented by matlab and the paper. One of the best

Shape Context Demo 
Serge Belongie ([e-mail])
Sept. 4, 2000


1. Information on Chui and Rangarajan data (for demo_1.m)

Each filename is of the form 
save_{fish,chinese,hand}_{def,noise,outlier}_m_n.mat
The value of m refers to the test number, n refers to the test example
number.  The range of m is 1-5 for deformation and outlier, 1-6 for
noise.  The range of n is 1-100.

Example: save_fish_outlier_3_2.mat
This corresponds to the "fish" shape, outlier test, test no. 3 (out of
5), corresponding to 100% outlier to data ratio, example no. 2.

Each file contains the model data (x1), the target data (y2a), and in
the case of the noise test, the target data before adding noise (y2).


2. Information on Digit Matching test (for demo_2.m)

This uses test data from Yann LeCun's MNIST database; see
http://www.research.att.com/~yann/ocr/mnist/ for details.