ImportError: cannot import name _postprocess_conv2d_output about densenet HOT 3 CLOSED

I find the answer. It is just the difference of keras. I used keras by fchollet. However this project is using farizrahman4u/keras-contrib. At last, I just go to this url:

https://github.com/farizrahman4u/keras-contrib/blob/master/keras_contrib/backend/tensorflow_backend.py

And copy _postprocess_conv2d_output and _preprocess_conv2d_input on my own. That is OK now!

from densenet.

Sorry for the delay. Got caught up in midsems. This should be fixed properly now for the Tensorflow backend

from densenet.

The enviroment that I run was:
tensoflow 1.10.0
keras 2.2.2

The problem occurred while I ran python train_coco.py:

Traceback (most recent call last):
File "train_coco.py", line 15, in
from models import *
File "/home/chun/src/Keras-FCN/models.py", line 6, in
from keras_contrib.applications import densenet
File "/home/chun/CHUN/anaconda3/envs/KFCN/lib/python3.6/site-packages/keras_contrib-1.2.1-py3.6.egg/keras_contrib/init.py", line 2, in
File "/home/chun/CHUN/anaconda3/envs/KFCN/lib/python3.6/site-packages/keras_contrib-1.2.1-py3.6.egg/keras_contrib/backend/init.py", line 14, in
File "/home/chun/CHUN/anaconda3/envs/KFCN/lib/python3.6/site-packages/keras_contrib-1.2.1-py3.6.egg/keras_contrib/backend/tensorflow_backend.py", line 10, in
ImportError: cannot import name '_postprocess_conv3d_output'

from densenet.