Skip to content

Dataset · iimmortall

Synthetic dataset for video HDR reconstruction from UE5

iimmortall/S2R-HDR-2

24,000 synthetic EXR frames for training high dynamic range video restoration models.

What's inside

A large synthetic image-to-image dataset rendered in Unreal Engine 5 via the XRFeitoria pipeline. It contains 1,000 motion sequences with 24 frames each, totaling 24,000 images at 1920x1080 resolution. All files are saved in EXR format with linear HDR values.

What you can do with it

  • Training models to convert SDR images and video to HDR
  • Training HDR reconstruction algorithms accounting for inter-frame motion
  • Testing augmentation and processing pipelines for linear EXR data
Tasks
Image → image
Downloads this month
31.6K
Likes
1

How to load it

from datasets import load_dataset

ds = load_dataset("iimmortall/S2R-HDR-2")