Closed codename1995 closed 5 years ago
attention_module.py
.... avg_pool = Lambda(lambda x: K.mean(x, axis=3, keepdims=True))(cbam_feature) assert avg_pool._keras_shape[-1] == 1 ....
The K.mean() seems not to work, cause the _avgpool have the same size with _cbamfeature.
Did you meet this problem?
This problem has benn solved. It is caused by using theano as backend. So, it can be fixed by using tensorflow. But I still haven't figured out the essential reason.
attention_module.py
The K.mean() seems not to work, cause the _avgpool have the same size with _cbamfeature.
Did you meet this problem?