Table 1 Overview of acronyms and definitions employed in the study.
Acronym | Definition |
|---|---|
CNN | Convolutional neural network |
FS | Feature selection |
RL | Reinforcement learning |
Off-policy PPO | Off-policy proximal policy optimization |
DE | Differential evolution |
CelebA | CelebFaces Attributes |
LFW | Labeled Faces in the Wild |
CASIA-WebFace | Chinese Academy of Sciences Institute of Automation WebFace |
VGGFace2 | Visual Geometry Group Face 2 |
DL | DL |
HO | Hyperparameter optimization |
GBM | Gradient boosting machines |
RF | Random forest |
DT | Decision tree |
LASSO | Least absolute shrinkage and selection operator |
CFS | Correlation-based FS |
SMOTE | Synthetic minority oversampling technique |
DRL | Deep reinforcement learning |
MLP | Multi-layer perceptrons |
HMS | Human mental search |
NLP | Natural language processing |
ML | Machine learning |
EU SPIRIT | Scalable Privacy-preserving Intelligence analysis for Resolving Identities in real-time |
MATLAB | Matrix laboratory |
FAHP | Fuzzy analytic hierarchy process |
CCTV | Closed-circuit television |
PCA | Principal component analysis |
DenseNet-169 | Dense convolutional network with 169 layers |
ReLU | Rectified linear unit |
MTCNN | Multi-task cascade neural network |
DNVPT | Deep neural vision processing techniques |
WNN | Wavelet neural network |
LSTM | Long short-term memory |
DNN | Deep neural network |
AE | Autoencoder |
YOLOv8 | You Only Look Once version 8 |
FaceNet | Facial Recognition Network |
VGGFace | Visual Geometry Group Face |
GhostFaceNets | A lightweight architecture designed for efficient edge-device deployment |
GJO-ANN | Golden Jackal optimized artificial neural network |
GAN | Generative adversarial network |
DS-AEAN | Dual-scale adaptive efficient attention network |
EAOA | Enhanced addax optimization algorithm |
TRPO | Trust region policy optimization |
KL | Kullback–Leibler |
GPBA | Generic population-based algorithm |
G-means | Geometric mean |
AUC | Area under the curve |
TP | True positive |
TN | True negative |
FP | False positive |
FN | False negative |
GB | Gigabyte |
RAM | Random-access memory |
CUDA | Compute unified device architecture |
cuDNN | Compute unified device architecture DNN library |
GPU | Graphics processing unit |
RTX | Ray tracing Texel extreme |
Ti | Titanium |
MISSL | Multi-input spatio-structural learning |
CNBA | Criminal network-based architecture |
DL-ACO | DL with ant colony optimization |
DCNN | Deep CNN |
FECNN | Facial expression-based CNN |
CLSTM | CNN-LSTM hybrid model |
QN-FR | Quantum networking face recognition |
YOLOv8-FI | YOLOv8-based forensic identification |
FVG-FR | FaceNet-VGG-GhostFaceNet recognition |
QWE-DNN | Quality-weighted embedding with DNN |
FacialCueNet | Facial cues network |
GAN-DSAEAN | GAN with dual-scale adaptive efficient attention network |
FLOP | Floating-point operations |
ITPS | Inference time per sample |
RTB | Real-time bidding |
FGSM | Fast gradient sign method |
mRMR | Minimum redundancy maximum relevance |
MI | Mutual information |
TSFS | Teacher‑student FS (TSFS) |
A‑SFS | Batch‑attention‑based self‑supervision FS (A‑SFS) |
ROC | Receiver operating characteristic |
PR | Precision-recall |
SHAP | Shapley additive explanations |
BO | Bayesian optimization |
SSA | Salp swarm algorithm |
COA | Cuckoo optimization algorithm |
FA | Firefly algorithm |
BA | Bat algorithm |
ABC | Artificial bee colony |
