python - How can i programmatically generate descriptors for an arbitrary data set? -
i analyzing set of pictures, want classify. classification done artificial neural network in supervised manner. have test set assigns each picture class.
what want generate lot of descriptors , pca on these , statistical analysis how descriptor able describe class of picture.
how can generate descriptors these pictures programatically? me in future classification problems too. let assume have enough computation power (100 core cluster) there libraries incorporate lot of descriptors images?
you can follow 2 approaches start:
- feature based, using methods such sift or gist followed called bag of words approach. vlfeat site contains excellent demo of this.
- deploying deeplearning algorithms, such sparse autoencoder learn basic features of dataset can further used classification.
Comments
Post a Comment