Convert tensorflow dataset to pandas dataframe

show_samples.py

import pandas as pd
import tensorflow as tf


def convert_tf_to_pd(ds, limit=32):
    """
    Read data from Tensorflow dataset to Pandas dataframe

    :param ds:
    :param limit:
    :return:
    """
    batch_iterator = ds.batch(limit).make_one_shot_iterator()
    with tf.Session() as sess:
        batch = batch_iterator.get_next()
        features_and_labels = sess.run(batch)
        samples = {
            **features_and_labels[0],
            'label': features_and_labels[1],
        }

    return pd.DataFrame.from_dict(samples)