Jump to content

Connect SuperML | Leeroopedia MCP: Equip your AI agents with best practices, code verification, and debugging knowledge. Powered by Leeroo — building Organizational Superintelligence. Contact us at founders@leeroo.com.

Implementation:Norrrrrrr lyn WAInjectBench Image Folder Data Schema

From Leeroopedia
Knowledge Sources
Domains Data_Engineering, Computer_Vision
Last Updated 2026-02-14 16:00 GMT

Overview

Concrete data schema for image-based prompt injection benchmark folders used by the WAInjectBench image detection pipeline.

Description

Image data is organized as numbered image files within scenario subfolders, nested under benign/ and malicious/ parent directories. The pipeline iterates subfolders via parent_path.iterdir() and counts files via folder_path.glob("*").

Usage

Prepare image data in this folder structure before running any image detector. The --data_dir argument (default "data/image") points to the root directory.

Code Reference

Source Location

  • Repository: WAInjectBench
  • File: data/image/ (data layout), main_image.py (L77-84 for folder discovery)

Signature

# Folder discovery in main_image.py
for folder_name in ["benign", "malicious"]:
    parent_path = data_dir / folder_name
    for sub_folder in parent_path.iterdir():
        if sub_folder.is_dir():
            res = process_folder(sub_folder, detector, detector_name,
                                 is_malicious=(folder_name == "malicious"))

Import

from pathlib import Path

I/O Contract

Inputs

Name Type Required Description
Image folders Directory Yes Subfolders containing numbered image files
Directory structure Filesystem Yes benign/ and malicious/ parent directories

Outputs

Name Type Description
Subfolder paths List[Path] Discovered scenario folders from iterdir()
is_malicious flag bool Derived from parent directory name
total_num int Count of files via glob("*")

Usage Examples

Directory Layout

data/image/
├── benign/
│   ├── normal_photos/
│   │   ├── 1.png
│   │   ├── 2.jpg
│   │   └── 3.png
│   └── documents/
│       ├── 1.png
│       └── 2.png
└── malicious/
    ├── text_overlay_attack/
    │   ├── 1.png
    │   ├── 2.png
    │   └── 3.png
    └── steganographic_attack/
        ├── 1.png
        └── 2.png

Related Pages

Implements Principle

Page Connections

Double-click a node to navigate. Hold to expand connections.
Principle
Implementation
Heuristic
Environment