---
pretty_name: CocoGlide
task_categories:
  - image-classification
  - image-segmentation
  - image-to-image
tags:
  - image-forgery-detection
  - image-manipulation-localization
  - synthetic-image-detection
  - diffusion-inpainting
  - GLIDE
  - COCO
---

# CocoGlide Dataset Card

## Dataset Description

This dataset is a reformatted version of **CocoGlide**, a local image forgery detection and localization dataset associated with the paper **TruFor: Leveraging All-Round Clues for Trustworthy Image Forgery Detection and Localization**.

CocoGlide is built from images in the MS-COCO validation set. The fake images are generated by applying **GLIDE-based text-guided inpainting** to local image regions. Each manipulated image is paired with a binary localization mask indicating the forged region.

The dataset can be used for both:

- **Image-level forgery detection**, where the goal is to classify whether an image is real or fake.
- **Pixel-level forgery localization**, where the goal is to predict the manipulated region.

## Dataset Structure

Each sample contains the following fields:

| Field | Type | Description |
|---|---|---|
| `key` | string | Sample identifier or image path |
| `image` | image | Input image |
| `mask` | image | Binary forgery localization mask |
| `label` | int | Image-level label, where `0` means real and `1` means fake |
| `prompt` | string or null | Text prompt/category used for GLIDE inpainting; `null` for real images |

## Label Definition

| Label | Meaning |
|---|---|
| `0` | Real image |
| `1` | GLIDE-inpainted fake image |

For real images, the corresponding mask should be treated as an all-zero mask.

## Intended Use

This dataset is intended for research on:

- Image forgery detection
- Image manipulation localization
- AI-generated image detection
- Diffusion-based inpainting forgery detection
- Robustness evaluation of forensic models

Typical evaluation settings include:

- Image-level metrics: Accuracy, AUC, AP, F1
- Pixel-level metrics: F1, IoU, AUC, localization AP

## Source Paper

This dataset should be cited through the following paper:

```bibtex
@InProceedings{Guillaro_2023_CVPR,
  author    = {Guillaro, Fabrizio and Cozzolino, Davide and Sud, Avneesh and Dufour, Nicholas and Verdoliva, Luisa},
  title     = {TruFor: Leveraging All-Round Clues for Trustworthy Image Forgery Detection and Localization},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  month     = {June},
  year      = {2023},
  pages     = {20606--20615}
}