chxy95/HDRTV1K
HDRTV1K HDRTV1K is a dataset for SDRTV-to-HDRTV conversion and related HDR imaging research, hosted on Hugging Face: Hugging Face Dataset: chxy95/HDRTV1K This dataset is associated with the HDRTVNet / HDRTVNet++ project line for SDRTV-to-HDRTV conversion. It contains HDR and SDR image data for training and testing, and is suitable for tasks such as HDR reconstruction, SDR-to-HDR translation, inverse tone mapping, and other low-level vision applications. According to the… See the full description on the dataset page: https://huggingface.co/datasets/chxy95/HDRTV1K.
HDRTV1K
HDRTV1K is a dataset for SDRTV-to-HDRTV conversion and related HDR imaging research, hosted on Hugging Face:
- Hugging Face Dataset:
chxy95/HDRTV1K
This dataset is associated with the HDRTVNet / HDRTVNet++ project line for SDRTV-to-HDRTV conversion. It contains HDR and SDR image data for training and testing, and is suitable for tasks such as HDR reconstruction, SDR-to-HDR translation, inverse tone mapping, and other low-level vision applications. According to the Hugging Face repository page, the dataset modality is Image and the license is MIT. (github.com)
Related Links
Dataset
- Hugging Face:
chxy95/HDRTV1K
GitHub Repositories
- HDRTVNet (ICCV 2021):
https://github.com/chxy95/HDRTVNet - HDRTVNet-plus / HDRTVNet++ (extended version):
https://github.com/xiaom233/HDRTVNet-plus
Papers
- A New Journey From SDRTV to HDRTV ICCV 2021:
https://arxiv.org/abs/2108.07978
- Towards Efficient SDRTV-to-HDRTV by Learning from Image Formation TMM:
https://ieeexplore.ieee.org/document/11146495
Overview
HDRTV1K is designed for research on converting SDR TV content to HDR TV content. It provides paired or corresponding SDR and HDR image data derived from HDR10 videos. The dataset is useful for:
- SDRTV-to-HDRTV conversion
- HDR reconstruction
- SDR-to-HDR translation
- Tone mapping / inverse tone mapping
- HDR image enhancement
- Low-level vision research
According to the HDRTVNet repository, the dataset is built from 4K-resolution HDR10 videos and their SDR counterparts. The paper further describes HDRTV1K as a dataset containing 1235 training images and 117 testing images. (github.com)
Dataset Structure
Based on the current Hugging Face repository file tree, the dataset is organized as follows:
HDRTV1K/
├── train_hdr/ # HDR training images
│ ├── black_001.png
│ ├── black_002.png
│ ├── ...
│ └── ...
├── train_sdr/ # SDR training images
│ ├── black_001.png
│ ├── black_002.png
│ ├── ...
│ └── ...
└── test_set/
├── test_hdr/ # HDR test images
│ ├── 001.png
│ ├── 002.png
│ ├── ...
│ └── ...
└── test_sdr/ # SDR test images
├── 001.png
├── 002.png
├── ...
└── ...