How can your code be applied to a new dataset?

[anonymous-HFN-Net]

Is there any preprocessing required for Raw format data? (e.g., how to obtain .mat files and .csv files). Thanks a lot.

[Hashe037]

The light fields collected are based on camera angle versus world scattering coordinates, so some photogrammetry is needed to calculate these world coordinates and correct for distortion. For the actual unmixing performed in MS-BSS, this is not critical but instead needed for accurate NLOS reconstructions and residual measurements (Section 4.4 in paper). MS-BSS in general is more robust to calibration differences. For MS-CPA, this preprocessing is critical since it relies heavily on this calibration information.

For MS-BSS on new datasets, I recommend analyzing "important_functions/performJadeBss.m". Unless you have detailed setup information for the light field occlusion-based imaging with an edge occluder, you likely would want to modify the code that actually performs the NLOS reconstructions ("performLfieldRecon.m" in utils) and use your own calibrated code in place. Also, while we obtain the best results using JADE and our specific preconditioner, other unmixing algorithm and preconditioners can be chosen as well. If you want to apply MS-CPA to new datasets, more care would have to be put into the calibration or the algorithm itself should be modified to fit your specific dataset.

[anonymous-HFN-Net]

Thank you for your detailed response， your work has been very inspiring to me.