See-in-the-Dark (SID) is a dataset for training and benchmarking single-image processing of raw low-light images. The dataset contains 5094 raw short-exposure images, each with a corresponding long-exposure reference image. Multiple short-exposure imagescan correspond to the same long-exposure reference image. The number of distinct long-exposure reference images in SID is 424.

The dataset contains both indoor and outdoor images. Images were captured using two cameras: Sony alpha 7S II and Fujifilm X-T2. These cameras have different sensors: the Sony camera has a full-frame Bayer sensor and the Fuji camera has an APS-C X-Trans sensor.

The dataset is accompanied by a pipeline for processing low-light images. The pipeline is based on end-to-end training of a fully-convolutional network. The network operates directly on raw sensor data and replaces much of the traditional image processing pipeline, which tends to perform poorly on such data.

Related publications:

  • Chen Chen, Qifeng Chen, Jia Xu, Vladlen Koltun, "Learning to See in the Dark", IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018.