FNO-USTC

An FNO based model builded for speed ​​reconstruction in ultrasound CT imaging. View My Poster

Model

graph LR
    A[Input] --> B{Preprocessing}
    B --> |envelope & phase| C[entry1]
    E[position embedding1] --> C[entry1]
    B --> |high frequence & phase| D[entry2]
    F[position embedding2] --> D[entry2]
    C --> G(Mix)
    D --> G
    G --> H[FNO Layer]
    H --> |...|I[Output]

Loading

Data

The data used for training is generated by DP, which can be obtained in this competition and their upcoming papers.

Download the data and put it in the data folder.

Training

First, set up the conifguration in config.toml.

Then, run the following command to train the model.

python train.py

The training process will be saved in the runs folder and can be monitoring using tensorboard.

Checkpoints will be saved in the ckpt folder.

Inference

python inference.py -m <model_path> -k3 <300k_path> -k4 <400k_path> -k5 <500k_path> -o <output_path>

License

MIT