Open SizheAn opened 2 weeks ago
Hi, we haven't tried that yet. The model introduced in our paper was trained to generate high-quality LDR images. Besides, I guess that if feeding a low-quality LDR image of an HDR scene directly, there will be hallucinations in the output image, because this LDR image is missing a lot of information, so the MSCN as the conditional input is also missing information to constrain the generated contents.
Hey I'm wondering if this model can also be used for converting LDR to HDR images? Thanks a lot!