ValueError with pspnet.py

[nhabili]

Hi, when I run pspnet.py, I get a ValueError due to a negative dimension (output file with error
pspnet.txt
attached). I'm using virtualenv with python2.7, tensorflow version 1.3.0 and keras version 2.0.8. I'm also using the weight files that you've provided.

Any help would be greatly appreciated!

Thanks
Nariman

[jmtatsch]

For me building the network still works.
can you print the input shapes in:
--> 177 prev_layer = AveragePooling2D(kernel, strides=strides)(prev_layer)

[nhabili]

Shapes are:

kernel = (60, 60)
strides = (60, 60)
prev_layer = (?, 2048, 60, 1)

[jmtatsch]

For me that is:

(60, 60)
(60, 60)
(?, 60, 60, 2048)

are you using theano dim ordering?

[Vladkryvoruchko]

For me everything also works.
Whats is steps to reproduce? Maybe you've done modifications to the code? Or maybe code stands as a lib and you're using wrong dimensions order.

[nhabili]

The issue was that "image_data_format" in keras.json was set to "channels_first". I removed that line and I don't have that problem anymore. Thanks for your help.