Benchmark RadarBG-REAL
BG-REAL is a benchmark for background manipulation detection and localization in images. It contains 7,000 processed samples (6,000 public-data anchored, 1,000 synthetic) over six edit families with matched authentic controls, source-group splits, and quality control.
- Released
- 2026-07-28
- Readiness
- Paper only
- Primary field
- Robotics & Autonomous Systems
Why it matters
Existing image forensics benchmarks focus on object-centric manipulations, missing background edits. BG-REAL provides a targeted evaluation with matched controls to measure false-positive rates from re-encoding artifacts, exposing a shared shortcut risk across baselines.
Motivation
Background manipulation is a practical but under-specified image-forensics setting: the manipulated evidence can sit outside the salient foreground object, while many evaluations emphasize object-centric copy-move, splicing, or generic synthetic edits.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.