A research team from University of California San Diego and Microsoft proposes Micro-Factorized Convolution (MF-Conv), a novel approach that can deal with extremely low computational costs (4M–21M FLOPs) and achieves significant performance gains over state of the art models in the low FLOP regime.

Here is a quick read: UCSD & Microsoft Improve Image Recognition With Extremely Low FLOPs.

The paper MicroNet: Improving Image Recognition with Extremely Low FLOPs is on arXiv.

Source link