PaReLab

Pattern recognition lab, an image classification toolbox using Knn classifier and corss-validation.

1, Data:

BnuCampus images and annotations.

2, DbInit:

Reading imageset with single or multiple labels into standard interface.

3, FeatureLbp:

Local Binary Pattern & Local Phase Quantization, based on (http://www.cse.oulu.fi/MVG/Downloads/LBPSoftware).

4, FeatureColorDescriptors:

Wrapped Koen's code (http://koen.me/research/colordescriptors/).

5, FeatureAsift:

Based on Yu's code (http://www.cmap.polytechnique.fr/~yu/research/ASIFT/demo.html).

6, FeatureDctQuantZigzag:

Discrete Cosine Transform, including quantization and zigzag-scanning.

7, ScSPM, according to the code by Jianchao Yang @ NEC Research Lab America (Cupertino).

8, KNN classifier:

Divide the training and test images (using cross validation), extract block features from multiple resolutions of each image, find nearest K samples to map their labels to test samples.