Workflow from Scientific Research

Recurrent-based architecture designs of DRCN [105] (upper left), DRRN [65] (upper right), and MemNet [92] (bottom). The DRCN applies the same convolutional layer multiple times. The DRNN extends this idea to recursive application of a residual block. The MemNet introduces a gate unit to distill most important features from all intermediate convolution results in a recursive manner.

DOI