dense_1 shape problem

[lucci17]

When I run the resnet module, I meet a problem with the shape of layer.Here is the output:
Traceback (most recent call last):
File "PredictFromCamera.py", line 89, in
video_capture(_args)
File "PredictFromCamera.py", line 26, in video_capture
model, labels_map, int_to_labels_map = load_model(args.data_root, args.model, args.snapshot)
File "/media/bongos/Seagate1/tensorflow/lego_detector/Predict.py", line 252, in load_model
model.load_weights(os.path.join(snapshot_path, snapshot), by_name=True)
File "/home/bongos/venv/lib/python3.5/site-packages/keras/engine/network.py", line 1163, in load_weights
reshape=reshape)
File "/home/bongos/venv/lib/python3.5/site-packages/keras/engine/saving.py", line 1149, in load_weights_from_hdf5_group_by_name
str(weight_values[i].shape) + '.')
ValueError: Layer #176 (named "dense_1"), weight <tf.Variable 'dense_1/kernel:0' shape=(100352, 38) dtype=float32_ref> has shape (100352, 38), but the saved weight has shape (2048, 38).

My environment is tensorflow-gpu 1.12.0, Keras 2.2.4, Keras-Applications 1.0.7, Keras-Preprocessing 1.0.9
Can you help me with the problem?

[kirill-sidorchuk]

Hi Lucci,
Thank you for reporting. The issue has something to do with global average pooling, which appears to be absent from the model. I will take a look.