GridMask Data Augmentation
January 14, 2020 ยท View on GitHub
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