titu1994
commented
5 years ago Thats odd. I can build EfficientNetB0 without any issue with no_top. Here's my output :
Downloading data from https://github.com/titu1994/keras-efficientnets/releases/download/v0.1/efficientnet-b0_notop.h5
16719872/16717576 [==============================] - 16s 1us/step
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
input_1 (InputLayer) (None, None, None, 3 0
__________________________________________________________________________________________________
conv2d_1 (Conv2D) (None, None, None, 3 864 input_1[0][0]
__________________________________________________________________________________________________
batch_normalization_1 (BatchNor (None, None, None, 3 128 conv2d_1[0][0]
__________________________________________________________________________________________________
swish_1 (Swish) (None, None, None, 3 0 batch_normalization_1[0][0]
__________________________________________________________________________________________________
depthwise_conv2d_1 (DepthwiseCo (None, None, None, 3 288 swish_1[0][0]
__________________________________________________________________________________________________
batch_normalization_2 (BatchNor (None, None, None, 3 128 depthwise_conv2d_1[0][0]
__________________________________________________________________________________________________
swish_2 (Swish) (None, None, None, 3 0 batch_normalization_2[0][0]
__________________________________________________________________________________________________
lambda_1 (Lambda) (None, 1, 1, 32) 0 swish_2[0][0]
__________________________________________________________________________________________________
conv2d_2 (Conv2D) (None, 1, 1, 8) 264 lambda_1[0][0]
__________________________________________________________________________________________________
swish_3 (Swish) (None, 1, 1, 8) 0 conv2d_2[0][0]
__________________________________________________________________________________________________
...
conv2d_64 (Conv2D) (None, None, None, 3 368640 multiply_16[0][0]
__________________________________________________________________________________________________
batch_normalization_48 (BatchNo (None, None, None, 3 1280 conv2d_64[0][0]
__________________________________________________________________________________________________
conv2d_65 (Conv2D) (None, None, None, 1 409600 batch_normalization_48[0][0]
__________________________________________________________________________________________________
batch_normalization_49 (BatchNo (None, None, None, 1 5120 conv2d_65[0][0]
__________________________________________________________________________________________________
swish_49 (Swish) (None, None, None, 1 0 batch_normalization_49[0][0]
==================================================================================================
Total params: 4,049,564
Trainable params: 4,007,548
Non-trainable params: 42,016
__________________________________________________________________________________________________I truncated the vast majority of the model building script, but nevertheless, the weights are downloaded and the model built with weights without issues.
Are you changing the input shape to something other than 224x224x3 ?
ryzbaka
Hi, I want to apply transfer learning to the model to detect oil storage tanks by applying transfer learning.However, when I import the EfficientNetB0 with include top=False, I get the error 'Operands could not be broadcast together with shapes (None, None, 24) (None, None, 16)', is this because its trying to apply the imagenet weight to a model without the top layer? If so, could you provide the weights for the same?