SaID

This repository is the official implementation of the paper SaID: Simulation-aware Image Denoising Pre-trained Model for Cryo-EM Micrographs.

v0.1.0

Complete training, testing, and an example micrograph of SaID is available now.
The simulated dataset will be available soon.

1 Operating System

Ubuntu 18.04 or Centos 7 is preferred.

2 Requirements

Python >= 3.6.13
Pytorch >= 1.7.1
opencv-python 4.5.1
numpy 1.19.2
scikit-image 0.17.1
scikit-learn 0.24.2
mrcfile 1.3.0
torchvision >= 0.8.2

3 Test Data

Example real dataset can be found at:
Google Drive: https://drive.google.com/file/d/1ECWjwu7GO55oAQRxKSvF-sP8JMoh8nWr/view?usp=sharing

4 Usage

The directory is advised to build as follows

./SaID
./SaID/denoised_micrographs
./SaID/test_data
./SaID/pre_trained_models

Run training/testing script

sh run_test.sh

For detailed parameter settings, please run

python denoise_cmd.py -h

Acknowledgement

We sinceresly thank following work with their open-sourced code. Code is modified from following work:
Bepler, T., Kelley, K., Noble, A.J., Berger, B. Topaz-Denoise: general deep denoising models for cryoEM and cryoET. Nat Commun 11, 5208 (2020).