---
pretty_name: TAID Dataset
license: cc-by-4.0
tags:
- intrinsic-image-decomposition
- terrain
- synthetic
---

# TAID-Dataset

Terrain intrinsic decomposition dataset with 16,000 scenes and one row per scene.

## Columns and numeric spaces

- `Input`, `A`, `S`, `V`: 8-bit RGB PNG. Byte values represent linear values in
  `[0, 1]`, quantized as `round(clamp(x, 0, 1) * 255)`. No sRGB/gamma transfer
  function is applied.
- `D`: NumPy `.npy` bytes (`float32`, HWC RGB), in linear HDR space `[0, 5]`.
- `water_mask`: 8-bit one-hot RGB PNG (`R=water`, `G=terrain`, `B=sky`).
- `D_filename`: original-style filename for the corresponding NPY payload.

Load diffuse shading with:

```python
import io
import numpy as np
from datasets import load_dataset

row = load_dataset("ShunTatsukawa/TAID-Dataset", split="train")[0]
D = np.load(io.BytesIO(row["D"]), allow_pickle=False)
```

The source diffuse target used `log1p(D) / log1p(5)`. This repository stores
the inverse-transformed value `expm1(encoded * log1p(5))`; it is not log-space.

## License

This dataset is licensed under the Creative Commons Attribution 4.0
International License (CC BY 4.0).