Spectral Hashing of images for matching with hamming distance

Question

I am looking for a Python implementation of a Spectral Hashing or Semantic Hashing algorithm for images. Which works on hamming distances so that I can easily store it in a database like MySQL or PostgreSQL.

I have also read about using a GIST to determine descriptors of a image and then using Spectral Hashing, LSH or a RBM. However, I don't know how to implement something like this in Python.

So far I have found some LSH code in Python, but I don't know if it generates a binary hash for Hamming distances.

I have tried pHashing, however that doesn't work nicely for object recognition in images.

I don't want a large arrays of keypoint descriptors where you have to calculate the euclidean distance, which is slow or need special data structures to match to other images.

Thanks

Answer 1

You can try this package: https://github.com/wanji/hdidx. It also implements the Product Quantization algorithm proposed by Herve Jegou :)