---
license: cc-by-4.0
pretty_name: synthetic image-classification dataset
task_categories:
- image-classification
tags:
- image
- image-classification
- synthetic-data
- text-to-image
- image-to-image
- edge-impulse
size_categories:
- n<1K
---

# Synthetic Image-Classification Dataset

Synthetic image-classification dataset generated with stable diffusion
(zerogpu_sdxl_turbo) using **text-to-image from class names + short descriptions**.

## Classes

| Label | Images |
|---|---|
| `background` | 20 |
| `coffee-mug` | 20 |
| `lamp` | 20 |

## Layout

```text
train/<label>/<label>.<id>.jpg
test/<label>/<label>.<id>.jpg
metadata.csv
```

## Loading

```python
from datasets import load_dataset

ds = load_dataset("imagefolder", data_dir="coffee-lamp")
# or, once pushed to the Hub:
ds = load_dataset("eoinedge/coffee-lamp")
print(ds)
```

## Edge Impulse

Filenames use the `label.<id>.jpg` convention, so they upload directly:

```bash
edge-impulse-uploader --category training train/**/*.jpg
```

## Notes

Synthetic images are a bootstrap for image-classification models. Validate with
real captures from the target device's camera before deployment.