wenzhengchina / matching-in-the-dark

Matching in the Dark: A Dataset for Matching Image Pairs of Low-light Scenes (ICCV2021)

MID dataset [Homepage]

Wenzheng Song, Masanori Suganuma, Xing Liu, Noriyuki Shimobayashi, Daisuke Maruta, Takayuki Okatani.

+ Matching in the Dark: A Dataset for Matching Image Pairs of Low-light Scenes

[Introduction] This repository contains details about the MID (Matching in the Dark) dataset. The MID dataset was introduced as a benchmark for local descriptor evaluation challenge in extreme low-light conditions. This dataset also can be used for low-light Raw image-enhancing evaluation. See the paper for more details.

• The dataset contains diverse scenes consisting of 54 outdoor and 54 indoor scenes.

• For each scene, we provide one pair of groups of multiple RAW-format images captured from different two positions.

• In each group, there are 48 (6 shutter speeds × 8 ISO settings) underexposure images and one correspond long-exposure image.

• We provide ground truth relative camera pose for each scene obtained with long-exposure images.

[Samples] Here are example stereo image pairs (long exposure versions) of four indoor scenes and four outdoor scenes!

If there is a need to manually get the MID dataset, download and untar the following file:

• MID [460GB].

Citation

If you find these models useful for your resesarch, please cite with this bibtex.

matching-in-the-dark's Issues

Hello，can you provide baidu netdisk about MID dataset？

Stereo Rectification

Hi!
Thanks for making this dataset open-source which has amazing utility in real-world applications. Is it possible to give the code for rectification of these images? or intrinsic parameters matrices or stereo camera calibration checkerboard images?

Is it possible to get ground truth fundamental matrix?

hi, is it possible to get ground truth fundamental matrix for every dataset sequence?
thank you.