From April to December 2012, my temporary work contract was dedicated to nilearn development.
Briefly, Nilearn is an API that aim to make the use of scikit-learn easier for neuro-imaging, from dataset loading to visualisation. This includes:
- dataset downloading management
- handling of several file formats
- neuroimaging specific algorithms (Searchlight...)
- visualization tools (not yet developed)
A big effort has been made on documentation so I strongly advise you to read the documentation if you are interested in. You can also find a blogpost related to Brainhack 2013 in which we developped a nilearn example from Cameron Craddock spectral clustering paper.