@kkruups @prabodhw96 @cchen156 Hi, congrats for your work. In your paper and code, during training, input and corresponding ground truth are first scaled between [0-1], and then input are scaled by ratio but the ground truth is still between [0-1]. Have you tried not to scale input by ratio and keep input and gt both in range [0-1] to train the network? Will it get better result and what's your opinion? THX!

paper has sRGB input comparision, I modify the last two level conv10(12 channel ) to output 3channel for the rgb input image generated by rawpy. the training loss is down to 0.025, the result is not bad, but some image is not real. is this correct way?

The official website contains a low light image and its result when run on SID model (http://web.engr.illinois.edu/~cchen156/SID/examples/6s.html). I could not find the weights for this in the repository. Is it possible to upload the weights for when trained on iPhone 6s?

Hello,thank you for your help.Because my computer configuration is too low, RAM is only 4G, gpu memory 2G, although you provide sonny(12G) data set, but still a bit big for me and I can't download it .so could you provide training test data set within 2G. I just want to use a small amount of data to test the model. Thank you !

I just tried to train Sony dataset after converting the raw data into jpg image and found the result quite terrible. So are there any tricks when training 8bit data? P.S.: The channel number of input is 12, which results from packing each channel of rgb into four.

I downloaded the sony data of 22 GB, the training and testing programs were executed, the folder named 'result_sony' is created, with some files in it, but those files are not executable and there were no images to be seen, can you please tell me that was result was expected once the program is executed completely.

google pixel 3 and 4's night sight appear to have very good results. is it possible that they got inspired by this paper to build their night sight mode? if no, how could they have done it and even improved it. on a further note, is it possible to deploy this model into a phone and do the calculations locally?

Hi,

Thanks for your novel work. I have some questions. I noticed that you restrict the value of ground truth image around 0 to 1, then force the model to generate output image which is also around 0 to 1. I am confused that why not just use the original ground truth image as the label and produce the output image directly without multiplying 255. Is there any difference between these two ways? Could you please help answer this question?

Thanks,

Hi! I was able to run the examples for Sony raw data and the results are amazing. I would like to experiment with Canon and Nikon raws but with Canon's raws the model seems to never converge. I am able to run it but i don't get any output. I modified the test_Sony file changing the ARW extension in CR2 (Canon's raw). Any advice?

Thank you!

Some important links in the README file are broken, for example a link to the paper shows 404. It seems particularly everything under http://cchen156.web.engr.illinois.edu/ is broken.

Hi,

Thanks for your great work. Since I didn't find any related information in the paper, I'd like to ask authors some questions about the Sony setting.

1. The Sony camera A7s2 supports silent shutter, or electronic shutter. Was the setting always on when the dataset is collected? Or the setting was sometimes on? Or all images were captured with the mechanical shutter?

2. Have you tried to estimate the black level or always used the black level embedded in the file (512)

why train data only use "00", such as 00001_00_0.1s.ARW, 00002_00_0.1s.ARW. other train data with "02", "03" not be used

Ubuntu20.04/Python2.7/Tensorflow1.15

Run script train_Sony.py, training starts but from the log it seems the Loss is not correct? Does anyone konw the issue?

333 154 Loss=nan Time=0.040 333 155 Loss=nan Time=0.041 333 156 Loss=nan Time=0.043 333 157 Loss=nan Time=0.036 333 158 Loss=nan Time=0.042 333 159 Loss=nan Time=0.040 333 160 Loss=nan Time=0.036 333 161 Loss=nan Time=0.042 334 1 Loss=nan Time=0.044 334 2 Loss=nan Time=0.042 334 3 Loss=nan Time=0.036

In SID dataset, for a particular scene there are majority of images which were taken using one exposure time. A minority of images are using different exposure times. Also the exposure time of the majority images vary among different scenes (1/30s or 0.04s or 0.1s).

Would you mind sharing how to determine the exposure time based on the scene when you prepare the SID dataset? Thanks a lot!