AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Reverse Engineering of Generative Models: Inferring Model Hyperparameters From Generated Images

Vishal Asnani, Xi Peng Yin, Tal Hassner, Xiaoming Liu

TPAMI 2023 · Image generation › Autoregressive and latent generation · approach: Generalizable fake image detection · 74 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add fingerprint estimation constraints (all four losses) (ablation)model parsing framework · network architecture continuousL1 error: 0.170 -> 0.149Improved
add magnitude loss to fingerprint estimation (ablation)model parsing framework · network architecture discreteF1 score: 0.674 -> 0.718Improved
add spectrum loss to fingerprint estimation (ablation)model parsing framework · network architectureL1 error and F1: L1 0.149 -> 0.149; F1 0.676 -> 0.718Mixed
add repetitive loss to fingerprint estimation (ablation)model parsing framework · loss function predictionF1 score: 0.794 -> 0.813Improved
add energy loss to fingerprint estimation (ablation)model parsing framework · network architecture continuousL1 error: 0.162 -> 0.149Improved
use cluster parser plus instance parser instead of one parser (ablation)model parsing framework · network architecture continuousL1 error: 0.161 -> 0.149Improved
use weighted cross-entropy instead of standard cross-entropy for discrete parameters (ablation)model parsing framework · discrete parameter confusion matricesclass prediction balanceImproved
average predictions over 10 test images instead of 1 (ablation)model parsing framework · network architecture continuousL1 error: 0.215 -> 0.151Improved
increase test images per GM from 100 to 500 (ablation)model parsing framework · network architecture continuousL1 error: 0.145 -> 0.146No clear effect
train on full 110 GMs instead of face GMs only (ablation)model parsing framework · face test GMs, network architecture continuousL1 error: 0.139 -> 0.118Improved
train on non-face GMs instead of face GMs when testing on face GMs (ablation)model parsing framework · face test GMs, network architecture continuousL1 error: 0.139 -> 0.213Hurt
use FEN plus parsing network instead of FEN alone (ablation)coordinated attack detector · coordinated misinformation attackAUC (%): 83.5 -> 87.3Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy