search

Linwei-Chen / LIS

IJCV2023 Instance Segmentation in the Dark
MIT License
78 stars 7 forks source link

Preprocess script for RAW files #7

Open wangchust opened 10 months ago

wangchust commented 10 months ago

The LIS is an excellent work and has contributed a lot to the community. I am new to this domain and have some questions.

  1. Is the raw image data used in LIS the same as that in LOD?
  2. Can the authors provide the processing script for converting the raw image data into .png files provided in this repo. I am quite interested in the influence of bit-depth.
Linwei-Chen commented 10 months ago

Thank you for your interest! The images in the LIS dataset are essentially the same as those in our previous work, LOD. The processing script is very simple: it first quantifies the raw image and then processes it into PNG format.

wangchust commented 10 months ago

Thank you for your reply! I am still a little confused about the raw format.

In other datasets like SID, raw images are in RGGB bayer format within one channel. I was using rawpy lib to read the raw file and demosaic it into RGB channels using cv2.cvtColor_BayerRGGB2RGB function. Then the so-called RAW-RGB can be quantified and saved as PNG format.

In LIS, do you also follow such pipelines?

Linwei-Chen commented 9 months ago

Yes, the raw images are also in RGGB Bayer format within one channel. Please note that, following SID, the G values are averaged from G1 and G2.

wangchust commented 9 months ago

Thank you! That's indeed helpful.

wangchust commented 8 months ago

Thank you for your interest! The images in the LIS dataset are essentially the same as those in our previous work, LOD. The processing script is very simple: it first quantifies the raw image and then processes it into PNG format.

By the way, about the quantifying process, can the authors provide the exact code or algorithm?