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.
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