Shape Mismatch Error Training voc-fcn16s during of copying parameters from pretrained model

[Pujan Paudel]

I tried training voc-fcn16s. The network configuration goes well, as the log goes : 

4368 net.cpp:255] Network initialization was done.

I0913 10:40:57.472801  4368 solver.cpp:56] Solver scaffolding done.

Then, the error comes as : 

Cannot copy param 0 weights from layer 'conv1_1'; shape mismatch. I am sure I have specified the exact  fcn16s-heavy-pascal.caffemodel on the solver.py file . 

What can I do further to debug this problem or move ahead with it? 

Any help would be great. 

Thanks