CODEWITHSUNDEEP
pythontensorflow
P Ujwal
P Ujwal

Reputation: 1

Tensorflow image resize does not work as expected

I am using the following code to resize an image, but am getting unexpected results. I am also attaching the output screenshots. Can someone explain why this is happening?

import cv2
import tensorflow as tf
import matplotlib.pyplot as plt

test_img = cv2.imread('deer12.jpeg')
test_img = cv2.cvtColor(test_img,cv2.COLOR_BGR2RGB)
plt.imshow(test_img, cmap=plt.cm.binary)
plt.show()
test_img=tf.image.resize(test_img,(32,32))
plt.imshow(test_img, cmap=plt.cm.binary)
plt.show()

Produces: img

Upvotes: 0

Views: 128

Answers (1)

Octav
Octav

Reputation: 485

Watch out for the messages which are displayed... They might be warnings.
"Clipping indput data..." - this suggests that you should normalize the image.

import cv2
import tensorflow as tf
import matplotlib.pyplot as plt

test_img = cv2.imread('deer12.jpeg')
test_img = cv2.cvtColor(test_img,cv2.COLOR_BGR2RGB)
plt.imshow(test_img, cmap=plt.cm.binary)
plt.show()

test_img = test_img / 255.0  # this one

test_img=tf.image.resize(test_img,[32,32])
plt.imshow(test_img, cmap=plt.cm.binary)
plt.show()

Upvotes: 1

Related Questions