pdra training stage

[BharathSeshadri]

Hello esr-nerf authors,

what happens in the pdra stage of the pipeline? and in your github page why do you run it for the esr-nerf dataset and not the dtu dataset? I ask because I'm trying to run it for my own dataset, but the pipeline fails at the pdra stage.

Best, Bharath

[ecrireme]

Thank you for your interest in our work.

The PDRA stage is designed to reconstruct emissive sources using multi-view images captured under two lighting conditions: with all emissive sources turned on and off. Since the DTU dataset does not contain emissive sources, the PDRA stage is not required for processing it.

If you would like to run the PDRA stage of ESR-NeRF on your own dataset, you’ll need to capture two sets of multi-view images—one with all emissive sources turned off and another with all of them turned on.
If your dataset does not include emissive sources to be reconstructed, running the PDRA stage is unnecessary.

[BharathSeshadri]

My dataset contains images with emissive sources of two ‘intensities’: high and low. Would it still work? If so, how can I organize the images in the dataset folders: dataset/test_nv, test_nvi, train?