Awesome

SSGD: A SMARTPHONE SCREEN GLASS DATASET FOR DEFECT DETECTION

part-1

Paper (ArXiv) | Results | Download(Baidu) | Download(TsinghuaCloud)

SSGD

For screen glass defect detection
2,504 annotated images and 3,914 defects
Seven types of defects: crack, broken, spot, scratch, light-leakage, blot, broken-membrane

Usage

This dataset is allowed for academic purposes only. We recommend that researchers use 5-fold cross-validation to conduct experiments on SSGD.

Dependency

We mainly depend pytorch, timm and mmdetection:

pip install torch==1.7.1+cu110 torchvision==0.8.2+cu110 torchaudio==0.7.2 -f https://download.pytorch.org/whl/torch_stable.html
pip install timm==0.3.2
pip install mmcv-full==1.3.17 -f https://download.openmmlab.com/mmcv/dist/cu110/torch1.7/index.html
pip install mmdet==2.25.0

After downloading the dataset, you should put it into ./data/defect_scree.

Citation

@inproceedings{SSGD,
  author    = {Haonan Han and
               Rui Yang and
               Shuyan Li and
               Runze Hu and
               Xiu Li},
  title     = {SSGD: A smartphone screen glass dataset for defect detection.},
  booktitle = {ICASSP},
  year      = {2023}
}