Abstract
KOREATECH-CGH provides RGBD–complex hologram pairs generated with a layer-based hologram method. Each sample bundles RGB, depth, amplitude, and phase at up to 2048 × 2048 resolution. The dataset targets machine learning–based computer-generated holography (ML-CGH) for hologram generation, upscaling, and related wave-optics tasks.
Directory Structure
root
├─test
│ ├─amp
│ │ └─*.exr
│ ├─depth
│ ├─img
│ └─phs
├─train
│ ├─amp
│ ├─depth
│ ├─img
│ └─phs
└─validation
├─amp
├─depth
├─img
└─phsHologram Configurations
| Config ID | Resolution | Pixel Pitch | Wavelengths (R,G,B) | Physical Extent (H × W × D) | Download |
|---|---|---|---|---|---|
| 256-3.6 | 256 × 256 | 3.6 μm | 638 nm, 532 nm, 450 nm | 0.9216 mm × 0.9216 mm × 10.1668 mm | 🤗 |
| 512-3.6 | 512 × 512 | 3.6 μm | 638 nm, 532 nm, 450 nm | 1.8432 mm × 1.8432 mm × 20.3336 mm | 🤗 |
| 1024-3.6 | 1024 × 1024 | 3.6 μm | 638 nm, 532 nm, 450 nm | 3.6864 mm × 3.6864 mm × 40.6672 mm | 🤗 |
| 2048-3.6 | 2048 × 2048 | 3.6 μm | 638 nm, 532 nm, 450 nm | 7.3728 mm × 7.3728 mm × 81.3344 mm | 🤗 |
| Comming soon | Comming soon | Comming soon | Comming soon | Comming soon | Comming soon |
Per-channel specs
| Data | Format | Channels | Precision | Scale |
|---|---|---|---|---|
| RGB | .exr | 3 | fp32 | 0-1 |
| Depth | .exr | 1 | fp32 | 0-1 |
| Amplitude | .exr | 3 | fp32 | Data dependent |
| Phase | .exr | 3 | fp32 | 0-1 |
Data Splits
| Split | Samples |
|---|---|
| Train | 5,000 |
| Validation | 500 |
| Test | 500 |
Acknowledgements
Supported by the National Research Foundation of Korea (NRF) through the Ministry of Education's Basic Science Research Program (Grant 2021R1I1A3048263, 50%) and by the Institute of Information and Communications Technology Planning and Evaluation (IITP) grant funded by the Korea Government (MSIT) (Grant 2019-0-00001, 50%).
Source 3D Models
RGB-D scenes were rendered from 3D meshes in the Google Scanned Objects dataset.
BibTeX
@article{LEE2026115636,
title = {A large-depth-range layer-based hologram dataset generation for machine learning-based 3D computer-generated holography},
journal = {Optics & Laser Technology},
volume = {203},
pages = {115636},
year = {2026},
issn = {0030-3992},
doi = {https://doi.org/10.1016/j.optlastec.2026.115636},
url = {https://www.sciencedirect.com/science/article/pii/S0030399226009874},
author = {Jaehong Lee and You Chan No and YoungWoo Kim and Duksu Kim},
keywords = {CGH(Computer-generated holography), Hologram, Machine-learning, Dataset, RGB-d, ML-CGH},
abstract = {Machine learning-based computer-generated holography (ML-CGH) has advanced rapidly in recent years, yet progress is constrained by the limited availability of high-quality, large-scale hologram datasets. To address this, we present KOREATECH-CGH, a publicly available dataset comprising 6000 pairs of RGB-D images and complex holograms across resolutions ranging from 256×256 to 2048×2048, with depth ranges extending to the theoretical limits of the angular spectrum method for wide 3D scene coverage. To improve hologram quality at large depth ranges, we introduce amplitude projection, a post-processing technique that replaces amplitude components of hologram wavefields at each depth layer while preserving phase. This approach enhances reconstruction fidelity, achieving 27.46 dB PSNR and 0.87 SSIM, surpassing a recent optimized silhouette-masking layer-based method by 3.86 dB and 0.09 SSIM, respectively. We further validate the utility of KOREATECH-CGH through experiments on hologram generation and super-resolution using state-of-the-art ML models, confirming its applicability for training and evaluating next-generation ML-CGH systems.}
}