A more parameter-efficient SOTA bottleneck! (2020/07)

CNN are great blablabla… Let's get to the point. SOTA for image classification on Imagenet is EfficientNet with 88.5% top 1 accuracy in 2020. In this article, I introduce a combination of EfficientNet and Efficient Channel Attention (ECA) to highlight the results of the ECA paper from Tianjin/Dalian/Harbin universities. MobileNetV2 is composed of multiple blocks which are called linear bottlenecks or inverted residuals (they're almost the same). Linear Bottleneck is a residual layer composed of one 1x1 convolution, followed by a 3x3 depthwise convolution, then finally a 1x1 convolution.