Balancing Foundation Models and Trait-Specific Learning for Forehead-Crease Recognition

Geetanjali Sharma; Aryan Jain; Dev Nagaich; Aditya Nigam; Raghavendra Ramachandra

cs.CVOct 21, 2026

Balancing Foundation Models and Trait-Specific Learning for Forehead-Crease Recognition

Authors: Geetanjali Sharma, Aryan Jain, Dev Nagaich, Aditya Nigam, Raghavendra Ramachandra

Organizations: Indian Institute of Technology Mandi, India · SAFE Center, Norwegian University of Science and Technology (NTNU), Gjøvik, Norway

Abstract

Biometric recognition often degrades under partial facial vis- ibility, where conventional facial cues are occluded. The forehead region remains consistently visible and contains distinctive crease patterns that can serve as a complementary biometric trait. We present a unified eval- uation of forehead-crease recognition across explainable zero-shot rea- soning, domain-adaptive learning, and trait-specific modeling. FIRE en- ables interpretable zero-shot verification using multimodal large language models, while pretrained face-recognition models are adapted to the fore- head domain through margin-based learning. Under open-set evaluation, the best zero-shot model GPT-4o+FIRE achieves 18.22% EER, while Edge-Forehead achieves 8.30% EER and 54.01% [email protected]% FMR. The results show that although MLLMs offer interpretable zero-shot verifi- cation, trait-specific learning remains more effective for robust forehead- crease biometric recognition.

Figures & tables

Figure 1: Forehead-crease verification pipeline using real and synthetic image pairs evaluated with MLLMs, FMs and trait-specific models.

Model	EER (%) $\downarrow$	TMR@FMR	TMR@FMR
		=0.1 (%) $\uparrow$	=0.01 (%) $\uparrow$
FH-V2(Real
Edge-Forehead(Our)	8.63	70.00	70.00
FH-SSTNet [ 15 ]	14.22	38.00	38.00
FHCVS [ 16 ]	21.56	32.00	32.00
CLIP [ 13 ]	24.11	14.00	6.00

Table 1: Identity verification results on the sampled FH-V2 verification subset (50 genuine and 450 impostor pairs) and synthetic subset. [ 16 ] .

Scenario	Round 1	Round 2	Round 3
Improved classification results and confidence	CLS: Non-Match Conf: 55% Dis Score: 65%	CLS: Match Conf: 70% Dis Score: 15%	CLS: Match Conf: 85% Dis Score: 20%
Improved match consistency with lower dissimilarity scores	CLS: Match Conf: 85% Dis Score: 15%	CLS: Match Conf: 85% Dis Score: 15%	CLS: Match Conf: 85% Dis Score: 10%
Improved forehead verification after iterative prompting	Unable to perform forensic analysis	CLS: NM Conf: 60% Dis Score: 75%	CLS: NM Conf: 75% Dis Score: 80%
Ambiguity due to conflicting or missing classifications	CLS: Match Conf: 75% Dis Score: 25%	Unable to perform forensic analysis or classify	CLS: NM Conf: 70% Dis Score: 60%

Table 2: Round-wise response comparison across three inference passes. CLS: classification; Conf: confidence; Dis: dissimilarity; M: match; NM: non-match.

Model	EER (%)	TMR (%) @ FMR (%) =
Model	EER (%)	0.1 (%)	0.01 (%)
FaceNet [ 14 ] + Arcface	20.65	11.72	2.49
FaceNet [ 14 ] + AdaFace	15.38	26.13	8.07
FaceNet [ 14 ] + CosFace	17.35	17.51	4.14
FaceNet [ 14 ] + MagFace	26.73	7.33	1.56
FaceNet [ 14 ] + ElasticFace	11.17	45.48	27.87

Table 3: Open-set verification performance of foundation, domain-adaptive, face-recognition, and trait-specific models on FH-V2 after fine-tuning on FH-V1.

Figure 2: DET curves for forehead-crease verification on the open-set FH-V2 dataset.

Biometric Identification

Biometric identification matches or verifies people and animals using distinctive physical, physiological, or behavioral signals.

See all 129 papers

Emergent Morphing Attack Detection in Open Multi-modal Large Language Models
Feb 17, 2026Marija Ivanovska, Vitomir ŠtrucImage Forgery Detection Face Presentation Attack Detection
Faculty of Electrical Engineering, University of Ljubljana, Tržaška cesta 25, Ljubljana, 1000, Slovenia
EXPL-FR: Explaining Face Recognition Models via Vision-Language Alignment
Aug 21, 2026Guray Ozgur, Mustafa Efe Tamyapar, Naser Damer +1Face Recognition Concept-Based Explanations
Fraunhofer IGD, Fraunhoferstr. 5, 64283 Darmstadt, Germany · TU Darmstadt, Karolinenpl. 5, 64289 Darmstadt, Germany
Unmasking Face Embeddings: Reading, Rendering and Naming with Foundation Models
Aug 31, 2026Fizza Rubab, Yiying Tong, Arun RossCross-Modal Alignment Face Recognition
Michigan State University

Feb 17, 2026J/K move · Enter open · S save

Emergent Morphing Attack Detection in Open Multi-modal Large Language Models

Marija Ivanovska, Vitomir Štruc

Faculty of Electrical Engineering, University of Ljubljana, Tržaška cesta 25, Ljubljana, 1000, Slovenia

Image Forgery Detection Face Presentation Attack Detection Multimodal Large Language Models Zero-Shot Learning Multimodal Model Evaluation

Open paper AskPDF arXiv

Face morphing attacks threaten biometric verification, yet most morphing attack detection (MAD) systems require task-specific training and generalize poorly to unseen attack types. Meanwhile, open-source multimodal large language models (MLLMs) have demonstrated strong visual-linguistic reasoning, but their potential in biometric forensics remains underexplored. In this paper, we present the first systematic zero-shot evaluation of open-source MLLMs for single-image MAD, using publicly available weights and a standardized, reproducible protocol. Across diverse morphing techniques, many MLLMs show non-trivial discriminative ability without any fine-tuning or domain adaptation, and LLaVA1.6-Mistral-7B achieves state-of-the-art performance, surpassing highly competitive task-specific MAD baselines by at least 23% in terms of equal error rate (EER). The results indicate that multimodal pretraining can implicitly encode fine-grained facial inconsistencies indicative of morphing artifacts, enabling zero-shot forensic sensitivity. Our findings position open-source MLLMs as reproducible, interpretable, and competitive foundations for biometric security and forensic image analysis. This emergent capability also highlights new opportunities to develop state-of-the-art MAD systems through targeted fine-tuning or lightweight adaptation, further improving accuracy and efficiency while preserving interpretability. To support future research, all code and evaluation protocols will be released upon publication.

Abstract

Biometric Identification

Explore similar work

Emergent Morphing Attack Detection in Open Multi-modal Large Language Models

EXPL-FR: Explaining Face Recognition Models via Vision-Language Alignment

Unmasking Face Embeddings: Reading, Rendering and Naming with Foundation Models