Keras-retinanet: Error when checking target: ValueError: expected regression to have shape (None, None, 4)

[francisbouchard]

I'm trying to train with my own dataset in a csv but I get this:
Epoch 1/20
Traceback (most recent call last):
File "examples/train_csv.py", line 117, in <module> keras.callbacks.ReduceLROnPlateau(monitor='loss', factor=0.1, patience=2, verbose=1, mode='auto', epsilon=0.0001, cooldown=0, min_lr=0), File "/home/Team09/Desktop/SOEN321/lib/python3.6/site-packages/keras/legacy/interfaces.py", line 87, in wrapper return func(*args, **kwargs) File "/home/Team09/Desktop/SOEN321/lib/python3.6/site-packages/keras/engine/training.py", line 2114, in fit_generator class_weight=class_weight) File "/home/Team09/Desktop/SOEN321/lib/python3.6/site-packages/keras/engine/training.py", line 1826, in train_on_batch check_batch_axis=True) File "/home/Team09/Desktop/SOEN321/lib/python3.6/site-packages/keras/engine/training.py", line 1411, in _standardize_user_data exception_prefix='target') File "/home/Team09/Desktop/SOEN321/lib/python3.6/site-packages/keras/engine/training.py", line 153, in _standardize_input_data str(array.shape)) ValueError: Error when checking target: expected regression to have shape (None, None, 4) but got array with shape (32, 67995, 5)

[hgaiser]

That is a bug in the latest version of Keras. Try using Keras 2.0.9.

Side note: could you leave this issue open? There are a lot of people experiencing this problem. Will help them to find it.

[francisbouchard]

Yes that worked for me.
Thanks for the help.

[zbsean]

yes, the error is keras 2.1.1 bug, It change to keras 2.0.9 , it is work

[de-vri-es]

Modified the title to help people find the issue quickly.

[danpeczek]

And if I can add something from my own experience: DON'T try to install keras in version lower than 2.0.9 - listen to authors. Just had an issue with keras 2.0.8 and line 99 in retinanet.py: outputs = keras.layers.Reshape((-1, 4), name='pyramid_regression_reshape')(outputs) - received following error: TypeError: int() argument must be a string, a bytes-like object or a number, not 'NoneType'

[hansoli68]

Is this bug fixed in keras master branch?

[hgaiser]

Yes.

[hgaiser]

This should be fixed with keras 2.1.2, closing this issue.