Spatial Context

We instantiate our strategy by designing an end-to-end learnable deep network, named as Dynamic Context Correspondence Network (DCCNet), for the semantic correspondence problem.