Dense-U-net-Tensorflow

Simple Tensorflow implementation of ["Dense-U-net: Dense encoder-decoder network for holographic imaging of 3D particle field" ]

Yufeng Wu, Jiachen Wu, Shangzhong Jin, Liangcai Cao, and Guofan Jin, "Dense-U-net: Dense encoder–decoder network for holographic imaging of 3D particle fields," Optics Communications 493, 126970 (2021).

Usage

Layer-oriented-algorithm-and-ground-true

• Used to generate data sets
• Change the corresponding storage address to your own data storage address

datasets

• The datasets is generated through Layer-oriented-algorithm-and-ground-true / creat_train_datasets.m.(use MATLAB)
• Layer-oriented-algorithm-and-ground-true/datasets used to store training datasets and test datasets
• For your dataset, put images like this:

├── dataset
   └── YOUR_DATASET_NAME (input data)
       ├── YOUR_DATASET_NAME 
           ├── YOUR_DATASET_NAME
       	        ├──xxx.tif (name, format doesnot matter)
	            ├──yyy.tif
	            └── ...
    └── YOUR_DATASET_NAME (ground true)
       ├── YOUR_DATASET_NAME 
           ├── YOUR_DATASET_NAME
       	        ├──xxx.tif (name, format doesnot matter)
	            ├──yyy.tif
	            └── ...

train

• Dense-U-net

Replace the data set address of "train_generator" in Dense-U-net /Dense-U-net.py with your own path. 
Change the location of the loss value of "plot_history" to your own address.
Set "is_train" under "if __name__ == '__main__':" to True.
Python Dense-U-net.py can train the Dense-U-net network

test

• Dense-U-net

Replace the data set address of "train_generator" in Dense-U-net /Dense-U-net.py with your own path. 
Change the location of the loss value of "plot_history" to your own address.
Set "is_train" under "if __name__ == '__main__':" to False.
"Cv2.imread" reads the data into its own address,
Python Dense-U-net.py can train the Dense-U-net network

particles_information_extraction

• Replace "cv2.imread" in the for loop with the address of the particle image you want to extract.
• python particles_information_extraction.py can extract particle information

Author

Yufeng Wu（[email protected]); Liangcai Cao([email protected])