Accuracy for Alexnet trained on places database

[Su]

Hi, 

I am new to caffe and I am trying to figure out how I can compute the accuracy of an alexnet trained on the places database.

I downloaded the test set so I have a directory containing some 14000 images. 

Searching though the web, I found that I need to convert these images to a database (lmdb) and add an accuracy layer to the network. Is there any documentation on this whole procedure?

What I am looking for is just a measure of performance of the model so that I can compare if I make changes to it. Any help would be appreciated.