search

jakeret / tf_unet

Generic U-Net Tensorflow implementation for image segmentation
GNU General Public License v3.0
1.9k stars 748 forks source link

Why my loss is always increasing? #161

Open pengzhipz opened 6 years ago

jakeret commented 6 years ago

We can't help you if you don't provide any kind of information

pengzhipz commented 6 years ago

@jakeret, Now I've solved this question, but the loss still can't converge, maybe for the rgb image? Is there any solution to slove this?

pengzhipz commented 6 years ago 
def _load_file(self, path, dtype=np.float32):
    img = cv2.imread(path, cv2.IMREAD_GRAYSCALE)
    img = img.resize((500, 376))
    return np.array(img, dtype)

help me transform my data to grayscale and resize, how to write the code ?