FPAP: A Folded Architecture for Efficient Computing of Convolutional Neural Networks | IEEE Conference Publication | IEEE Xplore