Back to datasets
Dataset assetOpen Source CommunityVideo DenoisingHigh‑Resolution Video

Recaptured Raw Video Denoising Dataset (ReCRVD Dataset)

We propose capturing noisy‑clean image pairs by sequentially displaying existing high‑resolution video frames on a 4K screen and re‑capturing the screen content. The dataset comprises 120 scenes, split into a training set (90 scenes) and a test set (30 scenes), covering five ISO levels ranging from 1,600 to 25,600.

Source
github
Created
May 28, 2024
Updated
Jun 16, 2024
Signals
128 views
Availability
Linked source ready
Overview

Dataset description and usage context

Dataset Overview

Dataset Name

  • Recaptured Raw Video Denoising Dataset (ReCRVD Dataset)

Content

  • Contains noisy‑clean video pairs for 120 scenes.
  • Data were obtained by sequentially displaying existing high‑resolution video frames on a 4K screen and re‑capturing the screen content.

Splits

  • Training set: 90 scenes
  • Test set: 30 scenes

Download

  • Google Drive: link
  • MEGA: link
  • Baidu Netdisk: link (extraction code: ogyw)

Usage

  • Visualization can be performed using the code in the "isp_matlab" folder or via the ISP module in RViDeNet.
Need downstream help?

Pair the dataset with AI analysis and content workflows.

Once the source passes your review, move straight into summarization, transformation, report drafting, or presentation generation with the JuheAI toolchain.

Explore AI studio