Cannot classify image using ResNet models

[Paul Zin]

I'm using pre-trained models from https://github.com/KaimingHe/deep-residual-networks

When I try to classify an image using this model I get an error:

base_conv_layer.cpp:189] Check failed: bottom[0]->num_axes() == first_spatial_axis + num_spatial_axes_ (3 vs. 4) bottom num_axes may not change.

How can I fix this?