from stanford dependency parser.
runs lexparser on sentences of each description file, feeds it to parse_sps2.py to save as sps2 format.
parses a given line into a set of sps2 phrases: (is (and beak black))
generates batches for training, evaluating and testing for nmn model, from cub data set and sps2 parses.
split from Mohamed on cub.
evaluates the task of zeroshot learning from nmn results.
calculates statistics based on a nmn logs json file.
spell checking script found online.
vocabulary used by spell_check.
code for generating negative examples (from cluster_info.pk) by olivia. (modified for minor fixes)
goes over the files parsed from sentence to sps2 format, and provides some statistics on the generated parse data set.
uses caffe to run vgg 16 forward pass on images and extract pool 5 layer features. has batching and uses opencv.
horizontally flips images on the given path, and save alongside original images in a given output path.