search

ZFTurbo / ZF_UNET_224_Pretrained_Model

Modification of convolutional neural net "UNET" for image segmentation in Keras framework
GNU General Public License v3.0
214 stars 64 forks source link

Tensorflow problem #4

Closed lawlite19 closed 6 years ago

lawlite19 commented 7 years ago

When I use tensorflow-1.3.0 as backend, I get this kind of error:

builtins.ValueError: Dimension 2 in both shapes must be equal, but are 3 and 32 for 'Assign' (op: 'Assign') with input shapes: [3,3,3,32], [3,3,32,3].
ZFTurbo commented 7 years ago

Can you please post full error message as well with line where error appears?

lawlite19 commented 7 years ago

Sorry, I made a mistake.

lawlite19 commented 7 years ago

Hi, could you send the pretrained weights for buildings to me. I notice DSTL competition has the task. Thank you very much. lawlitewang@gmail.com

ZFTurbo commented 6 years ago

Weights in DSTL competition was made for model with 20 input channels. They won't fit in current model. I recommend to train from scratch. UNET usually converges very well.

lawlite19 commented 6 years ago

Ok, thanks for your reply.