---
license: other
task_categories:
- image-classification
tags:
- ai-generated-image-detection
- deepfake-detection
- wildfake
dataset_info:
  features:
  - name: image_bytes
    dtype: binary
  - name: split
    dtype: string
  - name: group
    dtype: string
  - name: category
    dtype: string
  - name: source_zip
    dtype: string
  - name: source_path
    dtype: string
  - name: width
    dtype: int32
  - name: height
    dtype: int32
  - name: condition
    dtype: string
  - name: family
    dtype: string
  - name: order
    dtype: string
  splits:
  - name: train
    num_examples: 30000
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

# WildFake-Sample

A 30,000-image sample of **WildFake** (Hao et al., AAAI 2025,
[arXiv:2402.11843](https://arxiv.org/abs/2402.11843);
[original dataset](https://modelscope.cn/datasets/hy2628982280/WildFake)),
covering generators and real-image sources outside DDA/SID — a held-out
generalization slice, not a copy of the full ~3.6M-image dataset. All credit
for the images goes to WildFake's original authors. Built for
[Buxt-Codes/AIGI-Detection](https://github.com/Buxt-Codes/AIGI-Detection)
(branch `LoRC-PC`) — see that repo's `HANDOFF.md` for the evaluation
methodology and results.

## Composition

- **Fake (19,500):** 750 × 26 generators — GANs (BigGAN, StyleGAN, StarGAN,
  DF-GAN, GALIP, GigaGAN), non-SD diffusion (ADM, DDPM, DDIM, Imagen, VQDM,
  DALL-E 2/3, Midjourney v4/v5), SD-family (SDXL, OriginalSD, ControlNet,
  LoRA, LyCORIS, 2x personalized), other (MAGE, VQGAN, VQVAE, MAE).
- **Real (10,500):** 1,750 × 6 sources — LAION-5B, ImageNet, LSUN-Church,
  FFHQ, AFHQ, CelebA-HQ.

## Files

`data/train-*.parquet` — one row per image. `manifest.csv`/`.json` and
`transform_plan.csv`/`.json` — the same metadata as plain CSV/JSON.

**Columns:** `image_bytes` (raw file bytes, undecoded — decode with
`Image.open(io.BytesIO(row["image_bytes"]))`), `split`, `group`, `category`,
`source_zip`/`source_path`, `width`/`height`, `condition` (one of 14
transform-battery conditions, assigned round-robin per group, stratified by
resolution), `family`, `order` (always `transform_first`: condition applied,
then a final standardizing JPEG q=96 pass).

Full build methodology (HTTP range-request sampling, transform-assignment
algorithm, source code): see the GitHub repo linked above.