Created a function that batches a tf-record in order and apply data augmentation

Updated the LFW to tfrecord script to the new feature format.

Mind to review that (just few lines of code)?

Thanks

@tiago.pereira what is the difference between this and the generic one that we have?

Are you talking about what?

the script

We have this: https://gitlab.idiap.ch/bob/bob.learn.tensorflow/blob/b9c51932996e6577f68f8582f574e9fa06855d89/bob/learn/tensorflow/script/db_to_tfrecords.py

In order to reproduce the original LFW protocol, our database interface has an entity called Pairs, where a pair of samples (either from the same client or not) are compared. This script interacts directly with that.

Here is how you can convert this into a configuration for our main script:

import bob.db.lfw
import bob.io.base
import bob.io.image
import six

data_path = 'data_path'
output = 'output_file'
extension = 'extension'
protocol = 'protocol'
data_type = 'data-type'


def get_pairs(all_pairs, match=True):

  enroll = []
  probe = []
  for p in all_pairs:
    if p.is_match == match:
      enroll.append(p.enroll_file)
      probe.append(p.probe_file)

  return enroll, probe


# Loading LFW models
database = bob.db.lfw.Database()
enroll, probe = get_pairs(database.pairs(protocol=protocol), match=True)

client_ids = list(set([f.client_id for f in enroll] +
                      [f.client_id for f in probe]))
client_ids = dict(zip(client_ids, range(len(client_ids))))


def file_to_label(f):
  return client_ids[str(f.client_id)]


samples = (val for pair in six.moves.zip(enroll, probe) for val in pair)


def reader(f):
    path = f.make_path(data_path, extension)
    data = bob.io.image.to_matplotlib(
        bob.io.base.load(path)).astype(data_type)
    key = str(f.path).encode('utf-8')
    label = file_to_label(f)
    return (data, label, key)

Well, ok. I will remove that entry point script.

Thanks

added 1 commit

• ddbb1251 - Removing LFW tfrecord generator

Compare with previous version

Just dropped the LFW script

merged

mentioned in commit 2404d630

Thanks @tiago.pereira