Real-time bank document fraud detection: 6-signal CV+OCR fusion (ELA, Grad-CAM, MC Dropout, OCR-spatial IoU) → tiered risk scoring → deployed as AWS SageMaker inference API.
[CVPR 2026 Findings] Official repository for the paper "Leveraging Contrastive Learning for a Similarity-Guided Tampered Document Data Generation Pipeline".
Banking-optimized signature verification with Siamese CNN - Contrastive learning for forgery detection with high recall for fraud prevention
Detect hidden, residual, and machine-readable content in documents, images, and files.
AI-powered forensic analysis for detecting identity and educational document manipulation with explainable, multi-signal evidence.
Herramienta open source para auditoría forense de PDF: detecta revisiones ocultas por actualización incremental y redacciones defectuosas (texto extraíble bajo tachaduras), para peritaje informático.
An end-to-end document forensics pipeline for deepfake document detection and pixel-level tamper localization across 11,962 document and receipt samples.
Explainable, signature-aware PDF tampering and provenance analysis (pikepdf + pyHanko). Recovers earlier revisions, detects changes after signing, hidden content and active content.
An 8.8 MB document layout detector for damaged, scanned and photographed pages. Detects fifteen region types including form fields, and reports eight page-condition signals. Reaches 0.449 mAP@50-95 on degraded scans where a larger YOLOv8n baseline reaches 0.194.
Loan underwriting fraud checker. Cross-checks borrower docs against registries, bank flows, and land-use data — ELA forensics, Benford, and a consistency graph that shows where the story breaks.
Multi-modal AI framework for detecting synthetic & forged Indian identity documents (PAN, Aadhaar, Passport, UPI QR).
Document forgery detection — flags documents altered to imitate genuine originals. Built for ThinkRoot × Vortex'26 (NIT Trichy).
AI-powered document forgery & tampering detector — 7 forensic detectors + a trained two-stream CNN, fused into an AUTHENTIC / TAMPERED / AI-GENERATED verdict with a heatmap. FastAPI · PyTorch · React.
Font provenance reference — dated, source-verified timelines of bundled fonts and default-typeface changes across Windows, macOS, iOS, Android, Linux, office suites, PDF/OCR tools, and document libraries. For document dating and toolchain fingerprinting.
Deterministic layout reconstruction for messy documents. It reads PDF geometry (text operators, vector paths, the struct tree) and rebuilds columns, tables, figures, math and reading order into **GxDoc**, a typed intermediate structure that apps edit and export
Deep learning and Error Level Analysis for handwritten document forgery detection.
BorderShield AI -- SIH 2026 PS 26188: AI-based fake identity & document screening (Trust Ladder: crypto > rules > forensics/biometric)
Explainable tamper detection for PDFs and scanned documents: file structure, fonts, layout, pixels and arithmetic, with a plain-language reason for every finding.
Proofly — document and message verification with explainable fraud-risk analysis.
Forensic PDF auditor CLI. Detect edited, forged, or tampered PDFs with a 0-100 trust score. Local-first, no cloud, no uploads.
To associate your repository with the document-forensics topic, visit your repo's landing page and select "manage topics."