pyleargist

Pyleargist is a Python library that allows for easy computation of GIST global image descriptors, which are used to compare pictures based on their content. This makes it ideal for use in global scene recognition and categorization tasks. The theoretical definition of GIST image descriptors can be found on A. Torralba's page, while the source code for the C implementation is included in the lear_gist subfolder.

Pyleargist is licensed under the GPL, the same license as the original C project. To install pyleargist, you will need to install libfftw3 with development headers, python dev headers, gcc, the Python Imaging Library (PIL), and numpy. Once installed, the library can be tested locally by building it with "python setup.py build_ext -i" and exporting the PYTHONPATH to the "src" folder. Alternatively, you can build and install pyleargist system-wide by running "python setup.py build" and then "sudo python setup.py install."

To use pyleargist, simply import the library and the necessary modules, load an image using PIL, and compute its GIST descriptors using the leargist.color_gist() function. The resulting descriptors will have a fixed size of 960 by default, and can be used as an euclidian space to cluster images based on their content. This dimension can then be reduced to a 32 or 64 bit semantic hash using Locality Sensitive Hashing, Spectral Hashing, or Stacked Denoising Autoencoders.

If you're interested in sample implementations of picture semantic hashing with SDAs, you can check out the libsgd library. Pyleargist is a powerful and user-friendly library that can greatly improve the accuracy of global scene recognition and categorization tasks.