ForgeGuard AI uses state-of-the-art deep learning to detect manipulated and forged scientific images at the pixel level — protecting research integrity across journals, institutions, and forensic labs.
Trusted by researchers at
Every tool you need to verify image authenticity with unparalleled accuracy.
Generates a full probability heat map highlighting every manipulated region with sub-pixel resolution — not just a yes/no answer.
Leverages Facebook's DINOv2 Vision Transformer for rich, self-supervised features that generalize across image types and manipulation techniques.
TTA ensembles flipped variants of each image, dramatically reducing false positives and boosting confidence in ambiguous regions.
FastAPI backend with async processing. Submit an image and receive a detailed JSON result with metrics in under 3 seconds on average.
Gradient-enhanced adaptive masking dynamically computes thresholds per image — so no single setting ever limits detection quality.
Every analysis returns manipulated area in pixels, confidence mean, and adaptive threshold — exportable as JSON for audit trails.
Drag and drop any scientific image — microscopy, gel electrophoresis, MRI, satellite imagery — or use our REST API.
DINOv2 extracts rich visual features. Our decoder generates a pixel-level probability map, then TTA enhances reliability.
Receive a clear authentic/forged verdict with confidence metrics, manipulated pixel area, and a downloadable heat map overlay.
Upload any image and see ForgeGuard AI analyze it instantly (demo mode — sign up for full analysis).
Drag & drop an image here
or
PNG, JPG, TIFF up to 20MB
Want full pixel-level heat maps and batch analysis?
Sign Up Free →Start free. Scale as your research grows.
Perfect for individual researchers exploring the tool.
Built for labs and publication-ready forensic reports.
For journals, institutions, and large-scale forensic needs.