Learning-to-See-in-the-Dark icon indicating copy to clipboard operation
Learning-to-See-in-the-Dark copied to clipboard

Published 20 hours ago • verified publisher icon cchen156

16bit bayer filter

Open ghost0710 opened this issue 6 years ago • 6 comments

  1. Can I use a Sony network model for a 16bit bayer filter?
  2. Is it possible to share iphone 6s raw images(dng format) in paper?

ghost0710 avatar Jul 18 '18 00:07 ghost0710

Hey if you're looking to test on any RAW images check apps: RAW+ and MuseCam. They are free.

allendoss avatar Jul 18 '18 14:07 allendoss

  1. You can try. You'd better train your own model instead of using the pretrained model. It should work.
  2. Here it is: https://drive.google.com/file/d/1CWYmOcfh0gOvIrL40w2zGZJ5Rda_sylB/view?usp=sharing

cchen156 avatar Jul 20 '18 18:07 cchen156

@cchen156 Thank You your answer but I use your iphons6s.DNG , result is not good for your pretrained model (sony) your DNG file is black level 512? 14bit -bayer raw image? im = np.maximum(im - 512,0)/ (16383 - 512) #subtract the black level ratio is 300? 100? input_full = np.expand_dims(pack_raw(raw),axis=0) *ratio if you have iphone6s test code , share your code please.

ghost0710 avatar Jul 23 '18 00:07 ghost0710

Hi, I used 535 as the blacklevel for the iphone 6s image. The ratio is 100. You should able to reproduce the result now.

cchen156 avatar Jul 23 '18 17:07 cchen156

@cchen156
Thank You. The results are better than before, but still the results are worse than sony. Why? Can I get better results when training with Iphone6s low illumination Dataset? image

ghost0710 avatar Jul 24 '18 00:07 ghost0710

@ghost0710 Did you manage to improve the result? I am getting similar results using raw iphone photos.

l0stpenguin avatar Mar 04 '19 06:03 l0stpenguin