dvlab-research / GridMask

GridMask Data Augmentation

Introduction

This project is the code for implementing the GridMask augmentation for image classification and object detection. The full paper is available at: https://arxiv.org/abs/2001.04086.

Main Result

ImageNet

	Baseline	+ GridMask
ResNet-50	76.5	77.9
ResNet-101	78.0	79.1
ResNet-152	78.3	79.7

COCO2017

	Baseline	+ GridMask
FasterRCNN-R50-FPN	37.4	39.2
FasterRCNN-X101-FPN	41.2	42.6

You can find pretrained models to achieve the results in models