Tensorflow-data-loader

Reading data into tensorflow using tf.data function with configurations inclduing normalization, batch, epoch and three data augmentaion methods.

Example of usage and testing

import matplotlib.pyplot as plt
import tensorflow as tf
import data_loader

data_list = '/list/example.txt'
plt.ioff()

#Parse the images and masks, and return the data in batches, augmented optionally
data = data_loader.data_batch(data_list, augment=['flip_ud','flip_lr','rot90'], 
                                        normalize=True,batch_size=20, epoch = None)
#Get the image and mask op from the returned dataset
image_tensor, mask_tensor = data

with tf.Session() as sess:
    # Evaluate the tensors
    for i in range(1):
        image, mask = sess.run([image_tensor, mask_tensor])

        # Confirming everything is working by visualizing
        plt.figure('augmented image')
        plt.imshow(image[0, :, :, :])
        plt.figure('augmented mask')
        plt.imshow(mask[0, :, :,:])
        plt.show()
    # Do whatever you want now, like creating a feed dict and train your models

Reference

HasnainRaz/Tensorflow-input-pipeline