Scanning image by Caffe

[John Sachs]

Hello, guys! After reading literature on convolutional neural networks found an interesting approach. That we scan an image by a little window (from 4x4 to 64x64, for example) and it tells the class probability (or soft max meaning) for each window. Is it possible to do in Caffe and if so, can you describe the approach or give a link to an example.

Thanks, John.