The repository is the source code for paper "Non-line-of-Sight Imaging via Neural Transient Fields". [Paper]

The preprocessed data we use can be downloaded at [Google Drive] or [Baidu Netdisk] with password: netf

The raw data can be downloaded at Zaragoza NLOS synthetic dataset, f-k migration and Convolutional Approximations

We also provide MATLAB code 'zaragoza_preprocess.m' and 'fkdata_preprocess.m' to convert data from Zaragoza dataset and fk to fit NeTF for those who want to run NeTF at other scene.

Make sure that the dependcies in requirements.txt are installed, or they can be installed by

"pip install -r requirements.txt"

Make sure that data is place correctly like

NeTF_public
│   README.md
│   run_netf.py
│   ...
│
└───data
    │   fk_dragon_meas_180_min_256_preprocessed.mat
    │   ...
    │
    └───zaragozadataset
        │   zaragoza256_preprocessed.mat
        │   ...
 

Then run with preset settings:

"python run_netf.py --config configs/zaragoza_bunny.txt"

Different settings are stroaged at "./configs/".

Under preset settings, the training process takes around 24 hours on a single NVIDIA Tesla M40 GPU.

The final volume and slices from different view are stroaged at "./model"

The matlab script "show_result.m" is also provided to generate 2D images from different views and 3D density distribution.

And the comparision between predicted and measured histogram is stroaged at "./figure"

Please email [email protected] or [email protected] if you have any questions or suggestions.