Biometrics and Presentation Attack Detection
I study liveness detection, attack recognition, quality-difference enhancement, and system robustness for face, fingerprint, iris, and related biometric samples under complex conditions.
Biometrics and Image Quality Intelligence
Xinwei Liu is a Lecturer, Ph.D., master's supervisor, and CPC member. He received dual Ph.D. degrees in Computer Science from the Norwegian University of Science and Technology and the University of Caen Normandy. He has been selected for the Zhejiang Provincial University Leading Talent Development Program, the Ningbo Yongjiang Talent Program, and the Ningbo High-level Leading Talent Program. His research focuses on biometrics, computer vision, image quality assessment, and artificial intelligence.
- 27+
- SCI/EI Papers
- 10M+ RMB
- Project Funding
- 300+
- Citations
LFIW Dataset
13,180
Wild latent fingerprint images
01
Research
My research targets trustworthy visual perception in complex real-world scenarios, spanning image quality modeling, biometric presentation attack detection, urban target recognition, and industrial vision applications.
Image Quality Assessment and Enhancement
This direction studies no-reference image quality assessment, natural image enhancement, biometric image quality evaluation, and their impact on recognition performance in real-world vision systems.
Computer Vision and Intelligent Applications
Combining AI, object recognition, and industrial vision, this work covers real-scene 3D and video fusion, complex part handling, forklift pallet stacking, and obstacle detection.
02
Projects
I have led projects funded by the National Natural Science Foundation of China, Zhejiang Provincial Natural Science Foundation, Ningbo Natural Science Foundation, and Ningbo Yongjiang Talent Program, and have contributed to Ningbo key R&D and major science and technology projects.
Active
Key Technologies and Applications for Biometric Anti-spoofing Under Complex Conditions
This project addresses biometric security in complex environments by studying anti-spoofing, quality-difference enhancement, and application validation for trustworthy identity authentication.
- Source
- Ningbo Yongjiang Talent Program for Scientific and Technological Innovation
- Funding
- 1.00M RMB
- Period
- 2022.01 - 2026.12
- Role
- Principal Investigator
Active
Vision-based Forklift Pallet Auto-stacking and Obstacle Detection System
This industry project develops vision-based pallet auto-stacking, obstacle detection, and system integration for intelligent warehousing and industrial vehicle applications.
- Source
- Industry-sponsored Project
- Funding
- 0.51M RMB
- Period
- 2022.07 - 2026.12
- Role
- PI (1/2)
Completed
Biometric Liveness Detection and Optimization Under Complex Conditions Based on Image Quality Difference Enhancement
This project models image quality differences, enhances discriminative features, and optimizes performance for biometric liveness detection under complex conditions.
- Source
- NSFC Young Scientists Fund (Category C)
- Funding
- 0.30M RMB
- Period
- 2022.01 - 2024.12
- Role
- PI (1/1)
Completed
Key Anti-spoofing Technologies for Multimodal, Unconstrained, and Multi-attack Biometric Recognition
This project studies attack detection, feature representation, and anti-spoofing technologies for multimodal biometric systems under unconstrained scenarios and multiple attack types.
- Source
- Zhejiang Provincial Natural Science Foundation
- Funding
- 0.10M RMB
- Period
- 2022.01 - 2024.12
- Role
- PI (1/1)
Completed
Biometric Anti-spoofing Based on Image Quality Difference Amplification
This project studies quality-aware feature enhancement and detection models for biometric anti-spoofing through image quality difference amplification.
- Source
- Ningbo Natural Science Foundation
- Funding
- 0.05M RMB
- Period
- 2022.01 - 2023.12
- Role
- PI (1/3)
Completed
Key Technologies and Equipment for Fast and Precise Transfer of Highly Reflective Automotive Injection-molded Parts
This project studies fast transfer, visual inspection, and equipment systems for highly reflective automotive injection-molded parts.
- Source
- Ningbo Science and Innovation Yongjiang 2035 Key Technology R&D Program
- Funding
- 3.50M RMB
- Period
- 2023.07 - 2025.12
- Role
- Co-PI (2/22)
Completed
Real-scene 3D and Video Fusion with Urban Object Recognition
This project develops city-scale visual perception technologies for real-scene 3D, video fusion, and typical urban target recognition.
- Source
- Ningbo Major Science and Technology Project
- Funding
- 5.00M RMB
- Period
- 2022.03 - 2024.01
- Role
- Core Member (5/17)
03
Publications
I have published more than 27 papers in international journals and conferences, covering latent fingerprints, biometric image quality assessment, image enhancement, cross-domain recognition, and liveness detection.
Bridging Domains in Fingerprint Recognition With Quality-Normalized Fusion
IET Biometrics，SCI,中科院四区，DOI: 10.1049/bme2/9281903
2nd Latent in the Wild Fingerprint Recognition Competition
IEEE International Joint Conference on Biometrics，IJCB 2025，EI 收录，CCF-C，DOI: 10.1109/IJCB65343.2025.11410879
A Latent Fingerprint in the Wild Database
IEEE Transactions on Information Forensics and Security, SCI, CCF-A, CAS Q1 Top
Quality Matters: Boosting Face Presentation Attack Detection with Image Quality Metrics
IEEE Access, SCI, CAS Q3
Latent in the Wild Fingerprint Recognition Competition
IEEE International Joint Conference on Biometrics，EI，CCF-C
Survey of Natural Image Enhancement Techniques: Classification, Evaluation, Challenges, and Perspectives
Digital Signal Processing, SCI, CAS Q2
Performance Evaluation of No-reference Image Quality Metrics for Face Biometric Images
Journal of Electronic Imaging, SCI, CAS Q4
04
Teaching
I have long taught English-medium courses covering big data analysis, data visualization, web front-end design, and distributed application development, and have won provincial teaching competition awards.
Taught from 2021 to 2025 to 323 students, covering data analysis workflows, modeling methods, and applied practice.
Taught from 2020 to 2023 to 125 students, emphasizing data communication, interaction design, and visual analytics.
Taught from 2021 to 2022 to 234 students, covering HTML, CSS, JavaScript, and front-end engineering practice.
Taught from 2020 to 2021 to 32 students, focusing on distributed systems, application development, and engineering collaboration.
05
Latent Fingerprint Competition
The Latent in the Wild Fingerprint Recognition Competition is jointly organized by Xinwei Liu and Prof. Kiran Raja from NTNU. It establishes an open benchmark for latent fingerprint recognition and quality assessment in realistic complex scenarios.
From Real-world Scenes to an International Benchmark
Built on the LFIW database, the competition covers latent fingerprints from walls, iPad screens, aluminum foil, and other natural surfaces, pushing algorithms from controlled acquisition toward real forensic scenarios.
2026
3rd Edition: Ongoing
The 3rd edition at IJCB 2026 continues to focus on latent fingerprint recognition and quality assessment, organized by Xinwei Liu and Prof. Kiran Raja.
IJCB 2026
Started Mar 1
Results on May 25
Registration / Data Access
Algorithm Submission
Secure Evaluation
Results Release
View 2026 Information
The 3rd edition continues the dual-track design from 2025, emphasizing real latent fingerprint recognition, consistency between quality scores and matching performance, and reproducible cross-domain evaluation.
- Track 1：Latent Fingerprint Recognition。
- Track 2：Latent Fingerprint Quality Assessment。
- Evaluation: submitted algorithms are executed by the organizers in a secure environment with unified recognition and quality metrics.
2025
2nd Edition: Recognition + Quality Assessment
Held at IJCB 2025, the competition expanded to two tracks: latent fingerprint recognition and latent fingerprint quality assessment, with 12 registered teams and 8 valid submissions.
12 Registered Teams
8 Valid Submissions
2 Tracks
View 2025 Results
In Track 1, DRM LAT v02 achieved the best recognition performance, reducing EER from the 2024 best of 0.228 to 0.168 and improving AUC from 0.854 to 0.921.
- Track 1 Rank 1: DRM LAT v02, EER 0.168, AUC 0.921, FTER 1.29%.
- Track 1 Rank 2: VeriFinger v2025.1, EER 0.200, FTER 0%.
- Track 2 Rank 1: VeriFinger v2025.1, pAUC+ 0.0155 / 0.0126, FTER 0%.
- Track 2 Rank 2: AFQA, pAUC+ 0.0174 / 0.0146, FTER 3.39%.
2024
1st Latent Fingerprint Recognition Competition
Held at IJCB 2024, the first edition attracted 6 registered teams and 3 valid submissions from academia and industry.
6 Registered Teams
3 Valid Submissions
3 Countries
View 2024 Results
The evaluation set contains 72 subjects, 1,436 reference fingerprints, and 5,280 latent fingerprints across L-Wall, L-Ipad, and L-Alum surfaces.
- VeriFinger v13.1：EER 0.228，AUC 0.854，FMR10 0.275。
- LatentMinuComp v0：EER 0.290，AUC 0.791，FMR10 0.384。
- MarkIDNet: FTER 0%, but overall EER 0.481, indicating room for improvement.
06
News
Updates on research projects, publications, teaching awards, student mentoring, and academic activities.
The 3rd Latent in the Wild Fingerprint Recognition Competition is ongoing at IJCB 2026, continuing the dual tracks of latent recognition and quality assessment.
The paper "Bridging Domains in Fingerprint Recognition With Quality-Normalized Fusion" was published in IET Biometrics.
The 2nd Latent in the Wild competition was held at IJCB 2025 with recognition and quality assessment tracks, attracting 12 registered teams and 8 valid submissions.
The paper "2nd Latent in the Wild Fingerprint Recognition Competition" was accepted by IJCB 2025 and indexed by EI.
Awarded First Prize in the Engineering Group of the 14th Zhejiang Provincial Young University Teacher Teaching Competition.
The first Latent in the Wild competition was held at IJCB 2024, attracting 6 registered teams and establishing an open benchmark for real-world latent fingerprint recognition.
The paper "A Latent Fingerprint in the Wild Database" was published in IEEE Transactions on Information Forensics and Security.
Awarded Second Prize in the Engineering Group of the 12th Zhejiang Provincial Young University Teacher Teaching Competition and the 1st Ningbo University Bilingual Teaching Competition.
浙公网安备33021202004268号