Fully-convolutional-neural-network-FCN-for-semantic-segmentation-Tensorflow-implementation icon indicating copy to clipboard operation
Fully-convolutional-neural-network-FCN-for-semantic-segmentation-Tensorflow-implementation copied to clipboard

Published 20 hours ago • verified publisher icon sagieppel

Doesn't predict correctly

Open Djeane89 opened this issue 5 years ago • 4 comments

Hi,

Thank you so much for sharing your code. I have tried it.

The training and validation are run successfully. However, after run the Inference.py, the prediction result "/label" only shown black intensity in .png prediction image.

Please let me know where can be the problem?

Thanks

Djeane89 avatar May 17 '19 03:05 Djeane89

Is the value of the pixels is all zeros or is it just look black. The png image translates labels into 0-256 intensity values so if you have to say 5 labels the max intensity is 4 and the annotation mask will look dark even if the prediction was done correctly.

sagieppel avatar May 19 '19 17:05 sagieppel

looks like the value of the pixels is all zeros.

if it looks dark even if the prediction was done correctly, how can I check it?

Djeane89 avatar May 19 '19 18:05 Djeane89

im=cv2.imread("Prdicintion.png") # read predicted label map (if its not already loaded) print(im.max()) # print max value of label map

sagieppel avatar May 19 '19 18:05 sagieppel

I have checked it. The output is '0'

Djeane89 avatar May 21 '19 15:05 Djeane89