Adding fully connected layer to existing CaffeNet or Zeiler-Fergus Network

[Ronak Gupta]

When I add fully connected layer to existing CaffeNet (8 layers)/ ZF Net (8 layers), and train using the new network, with pre-trained caffenet/ZF model on ImageNet, I get following error: 

[ERROR] Train Caffe Model: Check failed: ShapeEquals(proto) shape mismatch (reshape not set)