CARPM-FIQA: Trajectory-Cumulative Scores Stabilize Face Image Quality Assessment
FraunhoferIGD · hf · 2026-10-06
Fraunhofer IGD tackles temporal instability in FR-integrated face image quality assessment—single-epoch quality estimates fluctuate as the feature space evolves—with CARPM-FIQA, which accumulates relative point margin scores (intra-class compactness vs inter-class separation) across the entire training trajectory.
Theoretically it reduces variance, improves MSE, and gains ranking stability with convergence guarantees. Controlled experiments on SynFIQA and ablations confirm consistent improvements. Against 12 FIQA methods on 8 benchmarks, 4 FR models, and two FMR thresholds, CARPM-FIQA(L) and (S) rank 4th and 6th of 17 by averaged pAUC-EDC/AUC-EDC, within a few percent of the best method. The authors argue temporal aggregation generalizes to training objectives that fluctuate with evolving representations.
More from Research
- CoNLL 2023 paper: instruction-tuned GPT models beat children on Theory of Mind tests — dioscuri · 2026-10-06
- How Shazam worked pre-AI: spectrogram peaks hashed into a hashtable lookup — deedydas · 2026-10-06
- AI formal verification hits reality: the halting problem blocks existing codebases — danbri · 2026-10-06
- GeoCR: one generalist model unifies cloud removal across 10 datasets, no fine-tuning needed — KAIST-VIClab · 2026-10-06
- Pi²: turning Wikipedia tables into long-context reasoning training data (COLM 2026) — tuvllms · 2026-10-06
- New RL simulation environment for port logistics built on real-world port data — yb2698 · 2026-10-06