search

neil454 / deep-motion

Use a DCNN to perform frame interpolation.
141 stars 29 forks source link

Error in get_unet_2 function #4

Closed YifeiAI closed 7 years ago

YifeiAI commented 7 years ago

When only run two lines: from FI_unet import get_unet_2 model = get_unet_2((6, 128, 384))

it shows error of: 11

when run:

from FI_unet import get_unet_2 model = get_unet()

it shows error of: 22

Note that I use: Python 3.5 Tensorflow 0.10.0 Keras 1.1.2

neil454 commented 7 years ago

For those who can't see the error images (I can only see them through email), here is the error:

ValueError: Filter must not be larger than the input: Filter: (3, 3) Input: (1, 32)

As for the solution, refer to this comment.

EDIT: Sorry I accidentally closed this issue. I reopened it just now. Let me know if that solution worked.

YifeiAI commented 7 years ago

Thank you neil! I added those two lines but it shows new error again when I run:

from FI_unet import get_unet_2 model = get_unet2((6, 128, 384))

649691890534176536 319038025832799103

YifeiAI commented 7 years ago

I solved this problem by change 124 line in conv10 = Convolution2D(int(input_shape[0]/2), 1, 1, activation='sigmoid')(bn9)

I think this error probably result in the difference between python2.x and python3.x