Work

A hardware acceleration Framework of Reconfigurable Edge Modules for Convolutional Neural Networks