# Day-to-Night Image Synthesis for Training Nighttime Neural ISPs Source: [Day-to-Night Image Synthesis for Training Nighttime Neural ISPs]() Collection index: [llms.txt]() Data updated: 2026-01-01T14:40:09.660Z Authors: Abhijith Punnappurath, Abdullah Abuolaim, Abdelrahman Abdelhamed, Alex Levinshtein, Michael S. Brown Published: 2022-06-06T16:15:45.000Z ## Abstract Many flagship smartphone cameras now use a dedicated neural image signal processor (ISP) to render noisy raw sensor images to the final processed output. Training nightmode ISP networks relies on large-scale datasets of image pairs with: (1) a noisy raw image captured with a short exposure and a high ISO gain; and (2) a ground truth low-noise raw image captured with a long exposure and low ISO that has been rendered through the ISP. Capturing such image pairs is tedious and time-consuming, requiring careful setup to ensure alignment between the image pairs. In addition, ground truth images are often prone to motion blur due to the long exposure. To address this problem, we propose a method that synthesizes nighttime images from daytime images. Daytime images are easy to capture, exhibit low-noise (even on smartphone cameras) and rarely suffer from motion blur. We outline a processing framework to convert daytime raw images to have the appearance of realistic nighttime raw images with different levels of noise. Our procedure allows us to easily produce aligned noisy and clean nighttime image pairs. We show the effectiveness of our synthesis framework by training neural ISPs for nightmode rendering. Furthermore, we demonstrate that using our synthetic nighttime images together with small amounts of real data (e.g., 5% to 10%) yields performance almost on par with training exclusively on real nighttime images. Our dataset and code are available at https://github.com/SamsungLabs/day-to-night. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Day-to-night image synthesis on Nighttime dataset No noise](): Day-to-night image synthesis; Nighttime dataset No noise; PSNR: 45.28. - [Day-to-night image synthesis on Nighttime ISO 1600](): Day-to-night image synthesis; Nighttime ISO 1600; PSNR: 37.41. - [Day-to-night image synthesis on Nighttime ISO 3200](): Day-to-night image synthesis; Nighttime ISO 3200; PSNR: 35.7. - [Neural ISP on Nighttime raw-sRGB noise-free (test)](): Neural ISP; Nighttime raw-sRGB noise-free (test); PSNR: 46.01. - [Neural ISP on Nighttime raw ISO 1600 (test)](): Neural ISP; Nighttime raw ISO 1600 (test); PSNR: 39.04. - [Neural ISP on Nighttime raw (ISO 3200) (test)](): Neural ISP; Nighttime raw (ISO 3200) (test); PSNR: 37.54. ## Original sources and implementation [Original paper]() [PDF]() [Implementation]()