tensorflow: Seg fault when using tf session with opencv 3

Hi,

We noticed that when we try to use tensorflow with opencv 3, it consistently seg faults and crashes. The commands are:

import cv2
import numpy as np
import tensorflow as tf

with tf.Session() as sess:
    img = cv2.imread('messi5.jpg', 0)
    rows, cols = img.shape
    M = np.float32([[1, 0, 100], [0, 1, 50]])
    dst = cv2.warpAffine(img, M, (cols, rows))
    cv2.imshow('img', dst)
    cv2.waitKey(0)
    cv2.destroyAllWindows()

Assuming you have a messi5.jpg in the folder (https://raw.githubusercontent.com/abidrahmank/OpenCV2-Python-Tutorials/master/data/messi5.jpg)

We see this issue when we use tensorflow 0.12 GPU enabled, opencv 3.2.0 and python 2.7.6, CUDA 8.0 and CuDNN 5.1.5 . We did not observe this issue with opencv version 2.4.13 or 2.4.9.

We will also filed a bug report on opencv (https://github.com/opencv/opencv/issues/8155)