To facilitate research in semi-supervised learning and cervical abnormal cell detection，we create an largely unlabeled cervical cytology dataset：
link：https://pan.baidu.com/s/1wWJY_TdyhtEI1cwQLBAPm
password：cmk2
Our dataset includes 2471 WSI（Whole Slide Images）. These WSI were obtained using two different slide scanning devices, with 1204 from scanner 1 and 1267 from scanner 2. By processing these WSI，hundreds of thousands of images can be obtained.
Our main experimental results and weights are as follows：
| Backbone | AP | AP@50 | AP@75 | weights |
|---|---|---|---|---|
| Resnet-50 | 26.9 | 48.7 | 27.0 | link |
| Resnet-101 | 28.1 | 50.4 | 28.2 | link |
| PVT | 28.8 | 50.8 | 28.9 | link |
| Swin-Transformer | 29.4 | 51.5 | 29.5 | link |
Our approach uses mmdetection, some modules and code refer to mmdetection(https://github.com/open-mmlab/mmdetection)