Cervical Cytopathology Image Refocusing via Multi-scale Attention Features and Domain Normalization

Environment

python 3.6
tensorflow-keras 1.13.1

Train

train_style_gray_rb.py (DNN)
t_clear_stage1.py (RFN_s1)
t_clear_stage2.p (RFN_s2)
t_clear_stage1_nocycle.py (RFN_s1_no_cycle)

Test

our_to_3D.py (DNN)
TestRefocus.py (RFN)

Configs

cf_style_gray_rb.py (DNN)
cf_clear_stage1.py (RFN_s1)
cf_clear_stage2.py (RFN_s2)

Data

We release the cytopathology refocusing dataset at Baidu Cloud (https://pan.baidu.com/s/11VsxE8n_uX4G2Nn2jnWWwg) with and an extracting code "refo".

Pretrained Nucleus Segmentation UNet Weights

"block380.h5" can be downloaded at the above Baidu Cloud URL

Defocus Map Esitimation

"Fast Defocus Map Estimation" Ding-Jie Chen, Hwann-Tzong Chen, and Long-Wen Chang ICIP 2016
https://github.com/vllab/fast_defocus_map

Email

[email protected] or [email protected]