Create a gist now

Instantly share code, notes, and snippets.

The mapreduce code to extract gist features from ImageNet images. To be used together with mincepie.
from mincepie import mapreducer, launcher
import gflags
import glob
import leargist
import numpy as np
import os
from PIL import Image
import uuid
# constant value
GIST_DIM = 960
GIST_DTYPE = np.float32
# gflags
gflags.DEFINE_string("input_folder", "",
"The folder that contains all input images, organized in synsets.")
gflags.DEFINE_string("output_folder", "",
"The folder that we write output features to")
FLAGS = gflags.FLAGS
def process_image(filename, max_size=256):
"""Takes an image name and computes the gist feature
im =
im.thumbnail((max_size, max_size), Image.ANTIALIAS)
return leargist.color_gist(im)
class PygistMapper(mapreducer.BasicMapper):
"""The ImageNet Compute mapper. The input value would be a synset name.
def map(self, key, value):
if type(value) is not str:
value = str(value)
files = glob.glob(os.path.join(FLAGS.input_folder, value, '*.JPEG'))
features = np.zeros((len(files), GIST_DIM), dtype = GIST_DTYPE)
for i, f in enumerate(files):
feat = process_image(f)
features[i] = feat
except Exception, e:
# we ignore the exception (maybe the image is corrupted or
# pygist has some bugs?)
print f, Exception, e
outname = str(uuid.uuid4()) + '.npy', outname), features)
yield value, outname
class PygistReducer(mapreducer.BasicReducer):
def reduce(self, key, values):
"""The Reducer basically renames the numpy file to the synset name
key: the synset name
value: the temporary name from map
os.rename(os.path.join(FLAGS.output_folder, values[0]),
os.path.join(FLAGS.output_folder, key + '.npy'))
return key
if __name__ == "__main__":
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment