Srameo / DNF

[CVPR 2023 Highlight] DNF: Decouple and Feedback Network for Seeing in the Dark
Other
139 stars 7 forks source link

why can't find the result #5

Closed Catherineasdzz closed 1 year ago

Catherineasdzz commented 1 year ago

When I run bash demos/images_process.sh successfully, I can only find .log and .yaml in runs/CVPR_DEMO/image_demo, and could not find results

Srameo commented 1 year ago

Hello! Could you please provide the runtime commands, the last 30 lines of the log file, and your specific runtime environment?

Catherineasdzz commented 1 year ago

Hello, thank you for your reply! I am using a GPU on an online server, and I have set up the environment according to the instructions you provided. And this is the command I use: image And the log is like that: image image

Srameo commented 1 year ago

It appears that our script did not find any images to process.

The -d parameter in the program only supports passing a folder path, not a file path. Additionally, the script currently only supports Sony's ".ARW" data format and the ".npy" format. If you have downloaded our Demo data, you can try passing the short_pack folder path and add the -a parameter at the end of the command (to indicate that the data format is ".npy"). You can find more details in readme.

The commands should be like:

bash demos/images_process.sh -p pretrained/model_best_sid_sony.pth -d /root/autodl-tmp/Campus/short_pack -r 100 -a 
Catherineasdzz commented 1 year ago

The problem has been solved! Thank you very much! And if I want to use my own images, do I still need to convert them to the npy format first?

Srameo commented 1 year ago

If your images are in Sony ARW format, there is no need to convert the data to the .npy format. However, for other data formats, you will need to convert the RAW data to the "RGBG 4-channel" format and save it as .npy files. You can refer to the specific conversion code in scripts/preprocess/preprocess_sid.py for the conversion process.

Catherineasdzz commented 1 year ago

Got it! Thank you very much!