12 May 2020 in Journal of Forensic Sciences
In this study, we aim to compare the performance of systems and forensic facial comparison experts in terms of likelihood ratio computation to assess the potential of the machine to support the human expert in the courtroom. In forensics, transparency in the methods is essential. Consequently, state‐of‐the‐art free software was preferred over commercial software. Three different open‐source automated systems chosen for their availability and clarity were as follows: OpenFace, SeetaFace, and FaceNet; all three based on convolutional neural networks that return a distance (OpenFace, FaceNet) or similarity (SeetaFace). The returned distance or similarity is converted to a likelihood ratio using three different distribution fits: parametric fit Weibull distribution, nonparametric fit kernel density estimation, and isotonic regression with pool adjacent violators algorithm. The results show that with low‐quality frontal images, automated systems have better performance to detect nonmatches than investigators: 100% of precision and specificity in confusion matrix against 89% and 86% obtained by investigators, but with good quality images forensic experts have better results. The rank correlation between investigators and software is around 80%. We conclude that the software can assist in reporting officers as it can do faster and more reliable comparisons with full‐frontal images, which can help the forensic expert in casework.
7 March 2022 in Forensic Science International
14 April 2023 in IEEE Journal of Selected Topics in Signal Processing
Identifying suspects from surveillance footage is a crucial task in forensic investigations, but it is often hindered by the variable conditions of observation and the large amounts of data. Face image quality (FIQ) is a metric that measures the usefulness of a face sample for facial recognition. Existing methods for automated FIQ assessment only provide a scalar value for quality, and do not indicate which factors are causing low quality. Additionally, these methods are computationally expensive, which makes current FIQ assessment methods unsuitable for large numbers of images. To address these issues, we introduce multi-task explainable quality networks (XQNets). XQNets provide both the quality value and the associated facial and environmental attributes, and automatically learn how each attribute contributes to the quality value during the training process. We also propose a dataset-agnostic quality pairing protocol to ensure that sample pairs are balanced across datasets and evaluations are fair. Our experiments on the LFW and SCface benchmarks show that our approach generalizes well across different datasets and outperforms state-of-the-art methods. Our method offers a fast, explainable approach to FIQ assessment, making it suitable for large-scale forensic applications.
29 February 2024 in Forensic Science International: Synergy
• Study tackles the problem of accurate likelihood ratios in surveillance video face recognition. • Proposes a frame selection method to enhance forensic face recognition by pairing frames based on quality and face attributes. • Optimal frame selection framework validated across various models and datasets. • As additional result, the study reveals that the use of super-resolution preprocessing via CodeFormer adversely affects the reliability of forensic face recognition.
9 April 2026 in European Journal on Criminal Policy and Research
29 September 2025 in IEEE Signal Processing Letters
Deep Neural Network (DNN) based person re-identification (ReID) systems achieve high accuracy but operate as black boxes, limiting their deployment in high-risk scenarios such as forensic analysis and security screening. We propose an inherently interpretable and privacy-aware ReID framework built upon the SOLIDER-ReID architecture. For that, first, we apply B-cosification to the Swin Transformer backbone, replacing key layers with B-cos transforms to enable faithful, contribution maps that reveal the model's internal decision-making processes. Second, we introduce a multi-task learning module with attribute-specific heads that predict semantically meaningful soft biometrics (e.g., clothing type, colors, accessories), enhancing both performance and interpretability. Experiments on the Market-1501 Attribute dataset and the DukeMTMC-attribute-ReID dataset demonstrate that our approach achieves competitive performance while delivering near real-time, faithful explanations and maintaining privacy.