StevenReitsma / kaggle-national-data-science-bowl

National Data Science Bowl competition entry for the Best Whale Wow team of the Radboud University Nijmegen. We ended 68th.
1 stars 0 forks source link

Save labels #13

Closed gzuidhof closed 9 years ago

gzuidhof commented 9 years ago

Save/load labels

Writes labels, in order of loaded files. Lables are saved as integers based upon a mapping of classnames in alphabetical order to natural numbers:

# Small part of the dictionary
{'echinoderm_larva_seastar_bipinnaria': 42, 'unknown_sticks': 119, 'tornaria_acorn_worm_larvae': 107, 'echinoderm_seacucumber_auricularia_larva': 44, 'ctenophore_lobate': 31, 'pteropod_triangle': 89, 'ctenophore_cestid': 28 ...

Save/load metadata

Writes metadata of processed data to file (mostly preprocessing options, such as the pad_size).

Loading is pretty straightforward:

In  [3]: import util

In  [4]: util.load_metadata()
Out [4]: 
{u'class_count': 121,
 u'image_size': 32,
 u'patch_size': 6,
 u'patches_per_image': 729,
 u'square_method': 'pad'}

Rename to follow conventions

Most identifiers such as variable names are renamed to follow agreed upon conventions.

Misc

- Fix builds on travis due to missing PIL library
- Unskip tests that failed due to this.
- Remove unused save function in util file
- Preprocess refactor to use **kwargs instead of a dozen arguments