GUARD-AI Checkpoints

Checkpoints used in the GUARD-AI benchmark.

Checkpoint format

import torch
ckpt = torch.load("wang2020/genimage_sd14_best.pth", map_location="cpu")
state_dict = ckpt["model"]      # detector backbone weights
threshold  = ckpt["threshold"]  # Youden-J calibrated decision threshold (retrained ckpts only)

All detectors output a single fake-logit; sigmoid(logit) >= threshold ⇒ "AI-generated". Architecture definitions live in each detector's original repository; the GUARD codebase vendors and wires them for evaluation.