search

elliottwu / DeepHDR

This is the implementation for Deep High Dynamic Range Imaging with Large Foreground Motions (ECCV'18)
MIT License
186 stars 38 forks source link

raw #14

Closed lr2427 closed 4 years ago

lr2427 commented 4 years ago

Hi, Thanks for your job. I have some questions . The paper describes that the radiometric calibration and gamma correction are needed if the LDR is not in RAW format, however I don't see them in the code. Is the input of the network is in rgb format? Otherwise, If the input image is raw format, how to use the network? is demosaicking needed to convert RAW to RGB?

elliottwu commented 4 years ago

Hi, The images in the dataset have been calibrated and gamma corrected. It's some detail that will only become important when recovering actual physical radiance. For perceptually plausible results, radiometric calibration is not necessary.

Demosiacking is needed. Ultimately, as long as the input preprocessing is consistent for training and testing, whichever input format you use should not really matter too much. If you are just using the trained model for inference, I would recommend feeding rgb images (they will be mapped to HDR domain using inverse gamma correction).