"UserWarning: Possibly corrupt EXIF data" while classifying images

Question

Below is my code for multi image classification. I am getting an error; I think it is because of loading and other for mismatching in dimension somewhere.

The error message starts where the code ends. Can anybody see the problem?

#importing necessary packages
import numpy as np
import matplotlib.pyplot as plt
import tensorflow as tf
from PIL import Image
import tflearn
import tensorflow as tf


import tflearn
#for writing text files

import glob
import os
import random
#reading images from a text file
from tflearn.data_utils import image_preloader
import math


IMAGE_FOLDER = 'C:/Users/kdeepshi/Desktop/PyforE/Face-Detection/Train'
TRAIN_DATA = 'C:/Users/kdeepshi/Desktop/PyforE/Face-Detection/training_data.txt'
TEST_DATA = 'C:/Users/kdeepshi/Desktop/PyforE/Face-Detection/test_data.txt'
VALIDATION_DATA = 'C:/Users/kdeepshi/Desktop/PyforE/Face-Detection/validation_data.txt'
train_proportion=0.7
test_proportion=0.2
validation_proportion=0.1


#read the image directories
filenames_image = os.listdir(IMAGE_FOLDER)
#shuffling the data is important otherwise the model will be fed with a single class data for a long time and
#network will not learn properly
random.shuffle(filenames_image)


#total number of images
total=len(filenames_image)
##  *****training data********
fr = open(TRAIN_DATA, 'w')
train_files=filenames_image[0: int(train_proportion*total)]
for filename in train_files:
    if filename[0:4] == 'Mark':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 0\n')
    elif filename[0:5] == 'lucas':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 1\n')
    elif filename[0:3] == 'Ann':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 2\n')
    elif filename[0:5] == 'Henry':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 3\n')
    elif filename[0:5] == 'Hanna':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 4\n')
    elif filename[0:4] == 'Jack':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 5\n')
    elif filename[0:5] == 'Harry':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 6\n')
    elif filename[0:3] == 'Lui':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 7\n')
    elif filename[0:6] == 'Karlos':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 8\n')
    elif filename[0:4] == 'Guan':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 9\n')

fr.close()
##  *****testing data********
fr = open(TEST_DATA, 'w')
test_files=filenames_image[int(math.ceil(train_proportion*total)):int(math.ceil((train_proportion+test_proportion)*total))]
for filename in test_files:
    if filename[0:4] == 'Mark':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 0\n')
    elif filename[0:5] == 'lucas':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 1\n')
    elif filename[0:3] == 'Ann':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 2\n')
    elif filename[0:5] == 'Henry':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 3\n')
    elif filename[0:5] == 'Hanna':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 4\n')
    elif filename[0:4] == 'Jack':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 5\n')
    elif filename[0:5] == 'Harry':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 6\n')
    elif filename[0:3] == 'Lui':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 7\n')
    elif filename[0:6] == 'Karlos':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 8\n')
    elif filename[0:4] == 'Guan':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 9\n')
fr.close()

##  *****validation data********
fr = open(VALIDATION_DATA, 'w')
valid_files=filenames_image[int(math.ceil((train_proportion+test_proportion)*total)):total]
for filename in valid_files:
    if filename[0:4] == 'Mark':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 0\n')
    elif filename[0:5] == 'lucas':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 1\n')
    elif filename[0:3] == 'Ann':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 2\n')
    elif filename[0:5] == 'Henry':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 3\n')
    elif filename[0:5] == 'Hanna':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 4\n')
    elif filename[0:4] == 'Jack':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 5\n')
    elif filename[0:5] == 'Harry':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 6\n')
    elif filename[0:3] == 'Lui':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 7\n')
    elif filename[0:6] == 'Karlos':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 8\n')
    elif filename[0:4] == 'Guan':
        fr.write(IMAGE_FOLDER + '/'+ filename + ' 9\n')
fr.close()

#Importing data
X_train, Y_train = image_preloader(TRAIN_DATA, image_shape=(56,56),mode='file', categorical_labels=True,normalize=True)
X_test, Y_test = image_preloader(TEST_DATA, image_shape=(56,56),mode='file', categorical_labels=True,normalize=True)
X_val, Y_val = image_preloader(VALIDATION_DATA, image_shape=(56,56),mode='file', categorical_labels=True,normalize=True)



print ("Dataset")
print ("Number of training images {}".format(len(X_train)))
print ("Number of testing images {}".format(len(X_test)))
print ("Number of validation images {}".format(len(X_val)))
print ("Shape of an image {}" .format(X_train[1].shape))
print ("Shape of label:{} ,number of classes: {}".format(Y_train[1].shape,len(Y_train[1])))


#Sample Image
plt.imshow(X_train[1])
plt.axis('off')
plt.title('Sample image with label {}'.format(Y_train[1]))
plt.show()
print(type(X_test))


#input image
x=tf.placeholder(tf.float32,shape=[None,56,56,3] , name='input_image')
#input class


y_=tf.placeholder(tf.float32,shape=[None, 10] , name='input_class')



input_layer=x
print("Hiiiiiiii No error till this point")


#convolutional layer 1 --convolution+RELU activation
conv_layer1=tflearn.layers.conv.conv_2d(input_layer, nb_filter=64, filter_size=5, strides=[1,1,1,1],
                                        padding='same', activation='relu', regularizer="L2", name='conv_layer_1')

#2x2 max pooling layer
out_layer1=tflearn.layers.conv.max_pool_2d(conv_layer1, 10)


#second convolutional layer
conv_layer2=tflearn.layers.conv.conv_2d(out_layer1, nb_filter=128, filter_size=5, strides=[1,1,1,1],
                                        padding='same', activation='relu',  regularizer="L2", name='conv_layer_2')
out_layer2=tflearn.layers.conv.max_pool_2d(conv_layer2, 10)
# third convolutional layer
conv_layer3=tflearn.layers.conv.conv_2d(out_layer2, nb_filter=128, filter_size=5, strides=[1,1,1,1],
                                        padding='same', activation='relu',  regularizer="L2", name='conv_layer_2')
out_layer3=tflearn.layers.conv.max_pool_2d(conv_layer3, 10)

#fully connected layer1
fcl= tflearn.layers.core.fully_connected(out_layer3, 4096, activation='relu' , name='FCL-1')
fcl_dropout_1 = tflearn.layers.core.dropout(fcl, 0.8)
#fully connected layer2
fc2= tflearn.layers.core.fully_connected(fcl_dropout_1, 4096, activation='relu' , name='FCL-2')
fcl_dropout_2 = tflearn.layers.core.dropout(fc2, 0.8)
#softmax layer output
y_predicted = tflearn.layers.core.fully_connected(fcl_dropout_2, 10, activation='softmax', name='output')
#loss function
cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(y_predicted+np.exp(-10)), reduction_indices=[1]))
#optimiser -
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
#calculating accuracy of our model
correct_prediction = tf.equal(tf.argmax(y_predicted,1), tf.argmax(y_,1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
# session parameters
sess = tf.InteractiveSession()
#initialising variables
init = tf.global_variables_initializer()
sess.run(init)
saver = tf.train.Saver()
save_path="C:/Users/kdeepshi/Desktop/PyforE/Face-Detection/mark2.ckpt"

g = tf.get_default_graph()



# every operations in our graph
[op.name for op in g.get_operations()]
epoch=5000

batch_size=20
previous_batch=0
for i in range(epoch):
    #batch wise training

    if previous_batch >= len(X_train) :
            previous_batch=0
    current_batch=previous_batch+batch_size
    x_input=X_train[previous_batch:current_batch]
    x_images=np.array(x_input)
    x_images=np.reshape(x_images,[batch_size,56,56,3])
    y_input=Y_train[previous_batch:current_batch]
    y_label=np.reshape(y_input,[batch_size,10])
    previous_batch=previous_batch+batch_size
    _,loss=sess.run([train_step, cross_entropy], feed_dict={x: x_images,y_: y_label})
    if i%500==0:
        n=50 #number of test samples
        X_test=np.array(X_test)
        x_test_images=np.reshape(X_test[0:n],[n,56,56,3])
        y_test_labels=np.reshape(Y_test[0:n],[n,10])
        Accuracy=sess.run(accuracy,feed_dict={x: x_test_images ,y_: y_test_labels})
        print("Iteration no :{} , Accuracy:{} , Loss : {}" .format(i,Accuracy,loss))
        saver.save(sess, save_path, global_step = i)
    elif i % 100 ==0:
        print("Iteration no :{} Loss : {}" .format(i,loss))
x_input=X_val
x_images=np.reshape(x_input,[len(X_val),56,56,3])
y_input=Y_val
y_label=np.reshape(y_input,[len(Y_val),10])

Accuracy_validation=sess.run(accuracy,feed_dict={x: x_images ,y_: y_label})
Accuracy_validation=round(Accuracy_validation*100,2)
print("Accuracy in the validation dataset: {} %".format(Accuracy_validation))


Test_FOLDER = 'C:/Users/kdeepshi/Desktop/PyforE/Face-Detection/Test'


filenames_image = os.listdir(Test_FOLDER)
total=len(filenames_image)
print(total)

test_files=filenames_image[0: int(total)]
for filename in test_files:

    marty=Image.open(Test_FOLDER+'/'+filename)
    marty_resize=marty.resize((56,56),Image.ANTIALIAS)
    marty_resize=np.array(marty_resize)
    marty_test=marty_resize/np.max(marty_resize).astype(float)
    marty_test=np.reshape(marty_test,[1,56,56,3])
    c=sess.run(y_predicted, feed_dict={x: marty_test})
    d= np.argmax(c)
#test your own images
#test_image=Image.open('/path to file')
#test_image= process_img(test_image)
#predicted_array= sess.run(y_predicted, feed_dict={x: test_image})
    #predicted_class= np.argmax(predicted_array)
    if d==0:

        print("This is Mark\n")
    elif d==1:

        print("This is lucas\n")
    elif d==2:
        print("This is Ann")
    elif d==3:

        print("This is Henrry\n")
    elif d==4:

        print("This is Hanna\n")
    elif d==5:
        print("This is Jack")

    elif d==6:

        print("This is Harry\n")
    elif d==7:
        print("This is Lui")
    elif d==8:

        print("This is Karlos\n")
    elif d==9:

        print("This is guan\n")

Here is the error:

C:\Users\kdeepshi\Desktop\PyforE\Face-Detection>multi.py

curses is not supported on this machine (please install/reinstall curses for an optimal experience)

Dataset

Number of training images 22

Number of testing images 6

Number of validation images 3

C:\ProgramData\Anaconda3\lib\site-packages\PIL\TiffImagePlugin.py:692: UserWarning: Possibly corrupt EXIF data. Expecting to read 12 bytes but only got 0. Skipping tag 270

"Skipping tag %s" % (size, len(data), tag))

C:\ProgramData\Anaconda3\lib\site-packages\PIL\TiffImagePlugin.py:692: UserWarning: Possibly corrupt EXIF data. Expecting to read 6 bytes but only got 0. Skipping tag 271

"Skipping tag %s" % (size, len(data), tag))

C:\ProgramData\Anaconda3\lib\site-packages\PIL\TiffImagePlugin.py:692: UserWarning: Possibly corrupt EXIF data. Expecting to read 8 bytes but only got 0. Skipping tag 272 "Skipping tag %s" % (size, len(data), tag))

C:\ProgramData\Anaconda3\lib\site-packages\PIL\TiffImagePlugin.py:692: UserWarning: Possibly corrupt EXIF data. Expecting to read 64 bytes but only got 0. Skipping tag 282 "Skipping tag %s" % (size, len(data), tag))

C:\ProgramData\Anaconda3\lib\site-packages\PIL\TiffImagePlugin.py:692: UserWarning: Possibly corrupt EXIF data. Expecting to read 64 bytes but only got 0. Skipping tag 283 "Skipping tag %s" % (size, len(data), tag))

C:\ProgramData\Anaconda3\lib\site-packages\PIL\TiffImagePlugin.py:692: UserWarning: Possibly corrupt EXIF data. Expecting to read 20 bytes but only got 0. Skipping tag 306 "Skipping tag %s" % (size, len(data), tag))

C:\ProgramData\Anaconda3\lib\site-packages\PIL\TiffImagePlugin.py:692: UserWarning: Possibly corrupt EXIF data. Expecting to read 24 bytes but only got 0. Skipping tag 529 "Skipping tag %s" % (size, len(data), tag))

C:\ProgramData\Anaconda3\lib\site-packages\PIL\TiffImagePlugin.py:692: UserWarning: Possibly corrupt EXIF data. Expecting to read 24 bytes but only got 0. Skipping tag 532 "Skipping tag %s" % (size, len(data), tag))

C:\ProgramData\Anaconda3\lib\site-packages\PIL\TiffImagePlugin.py:692: UserWarning: Possibly corrupt EXIF data. Expecting to read 40 bytes but only got 0. Skipping tag 33432 "Skipping tag %s" % (size, len(data), tag))

C:\ProgramData\Anaconda3\lib\site-packages\PIL\TiffImagePlugin.py:709: UserWarning: Corrupt EXIF data. Expecting to read 2 bytes but only got 0. warnings.warn(str(msg))

Shape of an image (56, 56, 3)

Shape of label:(10,) ,number of classes: 10

Answer 1

There is problem reading EXIF data from images in PIL module. I tend to think this is bug in PIL instead of images being corrupt. As this data is not needed for deep learning you can just cleans your files. To do that, download ExifTool and run following command:

exiftool -r -all= -ext JPEG D:\datasets\ImageNet\train

-r: recursive -all: Set all EXIF data to null -ext: File extensions

This works on WIndows as well as Linux.