Fig 1.
Structure of autoencoder.
Fig 2.
MLFF module diagram.
Fig 3.
UIIDD process based on MDAAE.
Fig 4.
GANs training process.
Fig 5.
Calculation process of self-attention module.
Fig 6.
UIIDD model framework integrating self-AM and GANs.
Fig 7.
UIIDD process based on autoencoder and GANs.
Table 1.
Test environment and specific configuration.
Fig 8.
RP curves of different methods.
Fig 9.
FPR and FNR at different thresholds.
Fig 10.
AUROC values and cross category generalization errors under different noise intensities.
Table 2.
Ablation study results on MVTec AD dataset.
Fig 11.
Reasoning time under different hardware platforms and input resolutions.
Fig 12.
GPU memory usage and IoU value distribution with different pixel sizes.
Fig 13.
FAR of dynamic changes during the training process and different lighting scenarios.
Fig 14.
Convergence time and detection efficiency during training.
Table 3.
Research methods and benchmark model performance comparison.