With multi-scale testing, we push the current best single model result to a new record of 60. 1% box AP and 52. 3% mask AP without using extra training data.
Ranked #2 on Instance Segmentation on COCO test-dev
Specifically, we present a novel block: Half Instance Normalization Block (HIN Block), to boost the performance of image restoration networks.
Ranked #1 on Single Image Deraining on Rain100H
We propose RepMLP, a multi-layer-perceptron-style neural network building block for image recognition, which is composed of a series of fully-connected (FC) layers.
Ranked #329 on Image Classification on ImageNet