Cross-Generator Generalization of CNN-Based AI Face Detectors: A Grad-CAM Region-Consistency Analysis
Fuente:
Zenodo
Gespeichert in:
| Hauptverfasser: | , |
|---|---|
| Format: | Recurso digital |
| Veröffentlicht: |
Zenodo
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866901533665263616 |
|---|---|
| author | Gargouri, Ismail Assif, Reda |
| author_facet | Gargouri, Ismail Assif, Reda |
| contents | Code, trained model checkpoints, per-image evaluation tables, region-consistency CSVs, perceptual-hash manifests, qualitative Grad-CAM panels, and the full paper draft for the cross-generator generalization study. Two CNN-based face detectors trained on CIPLAB GAN-era data and evaluated on a 1,531-image diffusion-generated out-of-distribution set against three pre-registered research questions. Headline finding: cross-architecture sign disagreement on every primary metric. |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_20262275 |
| institution | Zenodo |
| language | |
| publishDate | 2026 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | Cross-Generator Generalization of CNN-Based AI Face Detectors: A Grad-CAM Region-Consistency Analysis Gargouri, Ismail Assif, Reda deepfake detection cross-generator generalization Grad-CAM explainable artificial intelligence convolutional neural networks diffusion models region consistency out-of-distribution evaluation Code, trained model checkpoints, per-image evaluation tables, region-consistency CSVs, perceptual-hash manifests, qualitative Grad-CAM panels, and the full paper draft for the cross-generator generalization study. Two CNN-based face detectors trained on CIPLAB GAN-era data and evaluated on a 1,531-image diffusion-generated out-of-distribution set against three pre-registered research questions. Headline finding: cross-architecture sign disagreement on every primary metric. |
| title | Cross-Generator Generalization of CNN-Based AI Face Detectors: A Grad-CAM Region-Consistency Analysis |
| topic | deepfake detection cross-generator generalization Grad-CAM explainable artificial intelligence convolutional neural networks diffusion models region consistency out-of-distribution evaluation |
| url | https://doi.org/10.5281/zenodo.20262275 |