---
license: mit
pretty_name: Hybrid Helmholtz const_back raw data
tags:
- geophysics
- helmholtz
- wave-scattering
- operator-learning
---

# Hybrid Helmholtz const_back Raw Data

This dataset repository contains the raw NumPy arrays used by the paper
[Hybrid operator learning of wave scattering maps in high-contrast media](https://arxiv.org/abs/2602.11197).

[Github repo](https://github.com/davidMis/hybrid-operator-learning-of-wave-scattering-maps-in-high-contrast-media).

The repository is intentionally minimal. It contains only the four raw arrays
consumed by `scripts/prepare_data.py`; processed splits can be regenerated from
these files.

## Files

| Path | Dtype | Shape | Size |
| --- | --- | --- | --- |
| `const_back/velocity_sharp.npy` | `float32` | `[50000, 256, 256]` | 12.2 GB |
| `const_back/velocity_smooth.npy` | `float32` | `[50000, 256, 256]` | 12.2 GB |
| `const_back/pressure_sharp.npy` | `complex64` | `[50000, 256, 256]` | 24.4 GB |
| `const_back/pressure_smooth.npy` | `complex64` | `[50000, 256, 256]` | 24.4 GB |

## Usage

```bash
hf download dmis09/hybrid-operator-learning-of-wave-scattering-maps-in-high-contrast-media --repo-type dataset --local-dir data/raw
python scripts/prepare_data.py --raw-root data/raw --output-root data/processed --dataset const_back
```