demo.md

Try DNF!

• Try DNF!
  • Provided Data
  • Convert Your Own RAW Images to Numpy for Acceleration
  • Convert Your Own RAW Video
  • Tips for Scripts

Provided Data

Dataset	🔗 Download Links	Shot on	CFA Pattern
NKU Campus	[Google Drive][Baidu Cloud]	Synthetic	Bayer (RGGB)

After Downloading the final data folder should be organized like:

Campus
├── list.txt
├── long_pack      # Optional, if the dataset was synthetic
│   └── *.npy
├── long_post_int  # Optional, if the dataset was synthetic
│   └── *.npy
├── long_png       # Optional, if the dataset was synthetic
│   └── *.png
├── short_pack
│   └── *.npy
└── short_png
    └── *.png

Convert Your Own RAW Images to Numpy for Acceleration

If your raw data is with Sony ARW format, you could simply convert the data folder [DIR] by the following command:

Notice: your data should be saved in [DIR]/short folder.

python scripts/preprocess/preprocess_sid.py --data-path [DIR] --camera Sony --split short

Then the numpy array format of your own raw image could be found in [DIR]/short_pack.

Convert Your Own RAW Video

TBD

Tips for Scripts

All the shell scripts for demo could be found in demo/ folder.
Besides the images_process.sh and video_process.sh as described in README, the images_to_video.sh could transform a image sequenes into a video clip:

bash demos/images_to_video.sh -d [DIR] -s [SAVE_PATH] -f [FILE_NAME]
# [DIR] is the path to your images (jpg or png format).
# Your video clip could be found in [SAVE_PATH]/[FILE_NAME].mp4

# A simple example
bash demos/images_to_video.sh\
 -d dataset/Campus \
 -s runs/video \
 -f campus_short