Reproducibility package for: Evaluating Physics-Inspired Priors for Eye-Region Deepfake Detection: A Leakage-Aware Paired Multi-Seed Forensic Validation Study
Year: 2026
Type: software
Abstract: This repository contains the reproducibility package for Evaluating Physics-Inspired Priors for Eye-Region Deepfake Detection: A Leakage-Aware Paired Multi-Seed Forensic Validation Study. The package supports reproduction of the reported analyses, tables, and figures from frozen outputs. It includes per-seed prediction files, metric tables, statistical reporting outputs, source-aware validation tables, figure-generation inputs, manuscript table files, manifests with sha256 checksums, and available analysis notebooks and scripts. Raw media files and dataset-derived visual panels from third-party benchmark datasets are not redistributed. The package is intended to support forensic validation and reproducibility of the reported results, not to provide a new training dataset or a deployable detector.
Author: Promjiraprawat, Kamphol
Institution: Ramkhamhaeng University
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7167732002