def save_pedestrians(self, pedestrians):
train_pedestrians, val_pedestrians = ProcessData.save_pedestrians(
self, pedestrians)
def chunks(l, n):
for i in xrange(0, len(l), n):
yield l[i:i + n]
for pedestrians, data_file_func in [
(train_pedestrians, self.train_data_file),
(val_pedestrians, self.val_data_file)
]:
inputs = list(itertools.chain(*[p.inputs for p in pedestrians]))
outputs = list(itertools.chain(*[p.outputs for p in pedestrians]))
for i, (input_chunk, output_chunk) in enumerate(
zip(chunks(inputs, self.params['features_per_file']),
chunks(outputs, self.params['features_per_file']))):
fname = data_file_func(i)
features = [{
'fname':
ProcessDataNN._bytes_feature('{0}_{1}'.format(
os.path.splitext(os.path.basename(fname))[0], j)),
'input':
ProcessDataNN._floatlist_feature(np.ravel(input).tolist()),
'output':
ProcessDataNN._floatlist_feature(
np.ravel(output).tolist())
} for j, (input,
output) in enumerate(zip(input_chunk, output_chunk))]
self.save_tfrecord(fname, features)
return train_pedestrians, val_pedestrians
def __init__(self, folder, params):
ProcessData.__init__(self, folder, params)
def save_pedestrians(self, pedestrians):
train_pedestrians, val_pedestrians = ProcessData.save_pedestrians(
self, pedestrians)
return train_pedestrians, val_pedestrians