A5101950279- Face recognition and analysis
- Face and Expression Recognition
- Biometric Identification and Security
- AI in cancer detection
- Radiomics and Machine Learning in Medical Imaging
- Domain Adaptation and Few-Shot Learning
- Colorectal Cancer Screening and Detection
- Facial Nerve Paralysis Treatment and Research
- Video Surveillance and Tracking Methods
- Cell Image Analysis Techniques
- COVID-19 diagnosis using AI
- Medical Image Segmentation Techniques
- Digital Imaging for Blood Diseases
- Image Processing and 3D Reconstruction
- Gender Studies in Language
- Artificial Intelligence in Healthcare and Education
- Generative Adversarial Networks and Image Synthesis
- Energetic Materials and Combustion
- Pancreatitis Pathology and Treatment
- Speech Recognition and Synthesis
- Adversarial Robustness in Machine Learning
- Digital Media Forensic Detection
- Time Series Analysis and Forecasting
- Gait Recognition and Analysis
- Face Recognition and Perception
INESC TEC
2021-2025
Universidade do Porto
2021-2024
Instituto de Engenharia de Sistemas e Computadores Investigação e Desenvolvimento
2021-2023
Faculdade de Medicina do ABC
2018
Centro Hospitalar do Baixo Vouga
2018
Abstract Considering the profound transformation affecting pathology practice, we aimed to develop a scalable artificial intelligence (AI) system diagnose colorectal cancer from whole-slide images (WSI). For this, propose deep learning (DL) that learns weak labels, sampling strategy reduces number of training samples by factor six without compromising performance, an approach leverage small subset fully annotated samples, and prototype with explainable predictions, active features...
This paper presents a summary of the Masked Face Recognition Competitions (MFR) held within 2021 International Joint Conference on Biometrics (IJCB 2021). The competition attracted total 10 participating teams with valid submissions. affiliations these are diverse and associated academia industry in nine different countries. These successfully submitted 18 solutions. is designed to motivate solutions aiming at enhancing face recognition accuracy masked faces. Moreover, considered...
This paper presents a summary of the Competition on Face Morphing Attack Detection Based Privacy-aware Synthetic Training Data (SYN-MAD) held at 2022 In-ternational Joint Conference Biometrics (IJCB 2022). The competition attracted total 12 participating teams, both from academia and industry present in 11 differ-ent countries. In end, seven valid submissions were submitted by teams evaluated organizers. was to at-tract solutions that deal with detecting face morphing at-tacks while...
Colorectal cancer (CRC) diagnosis is based on samples obtained from biopsies, assessed in pathology laboratories. Due to population growth and ageing, as well better screening programs, the CRC incidence rate has been increasing, leading a higher workload for pathologists. In this sense, application of AI automatic diagnosis, particularly whole-slide images (WSI), utmost relevance, order assist professionals case triage review. work, we propose an interpretable semi-supervised approach...
The development of deep learning algorithms has extensively empowered humanity's task automatization capacity. However, the huge improvement in performance these models is highly correlated with their increasing level complexity, limiting usefulness human-oriented applications, which are usually deployed resource-constrained devices. This led to compression techniques that drastically reduce computational and memory costs without significant degradation. These compressed especially essential...
SARS-CoV-2 has presented direct and indirect challenges to the scientific community. One of most prominent advents from mandatory use face masks in a large number countries. Face recognition methods struggle perform identity verification with similar accuracy on masked unmasked individuals. It been shown that performance these drops considerably presence masks, especially if reference image is unmasked. We propose FocusFace, multi-task architecture uses contrastive learning be able...
The ongoing use of chemical warfare agents (CWA) as weapons destruction remains a dangerous threat in current times. In the aftermath CWA incident, it is vital to conduct thorough decontamination surrounding environment, encompassing all exposed items. At present time, systems continue rely heavily on bleach-based solutions due broad range action. limitations these include potential for material destruction, high toxicity, generation hazardous by-products, and necessity waste treatment prior...
Over the years, evolution of face recognition (FR) algorithms has been steep and accelerated by a myriad factors. Motivated unexpected elements found in real-world scenarios, researchers have investigated developed number methods for occluded (OFR). However, due to SarS-Cov2 pandemic, masked (MFR) research branched from OFR became hot urgent challenge. Due time data constraints, these models followed different novel approaches handle lower occlusions, i.e., masks. Hence, this study aims...
To accurately explore the anatomical organization of neural circuits in brain, it is crucial to map experimental brain data onto a standardized system coordinates. Studying 2D histological mouse slices remains standard procedure many laboratories. Mapping these challenging; due deformations, artifacts, and tilted angles introduced during preparation slicing process. In addition, analysis can be highly dependent on level expertise human operator. Here we propose computational tool for...
In the context of biometrics, matching confidence refers to that a given decision is correct. Since many biometric systems operate in critical decision-making processes, such as forensics investigations, accurately and reliably stating becomes high importance. Previous works on estimation can well differentiate between low confidence, but lack interpretability. Therefore, they do not provide accurate probabilistic estimates correctness decision. this work, we propose interpretable comparison...
As two sides of the same coin, causality and explainable artificial intelligence (xAI) were initially proposed developed with different goals. However, latter can only be complete when seen through lens framework. such, we propose a novel causality-inspired framework for xAI that creates an environment development approaches. To show its applicability, biometrics was used as case study. For this, have analysed 81 research papers on myriad biometric modalities tasks. We categorised each these...
The recent Covid-19 pandemic and the fact that wearing masks in public is now mandatory several countries, created challenges use of face recognition systems (FRS). In this work, we address challenge masked (MFR) focus on evaluating verification performance FRS when verifying vs unmasked faces compared to only faces. We propose a methodology combines traditional triplet loss mean squared error (MSE) intending improve robustness an MFR system masked-unmasked comparison mode. results obtained...
Abstract Cervical cancer is the fourth most common female worldwide and leading cause of cancer-related death in women. Nonetheless, it also among successfully preventable treatable types cancer, provided early identified properly managed. As such, detection pre-cancerous lesions crucial. These are detected squamous epithelium uterine cervix graded as low- or high-grade intraepithelial lesions, known LSIL HSIL, respectively. Due to their complex nature, this classification can become very...
This work summarizes the IJCB Occluded Face Recognition Competition 2022 (IJCB-OCFR-2022) embraced by International Joint Conference on Biometrics (IJCB 2022). OCFR-2022 attracted a total of 3 participating teams, from academia. Eventually, six valid submissions were submitted and then evaluated organizers. The competition was held to address challenge face recognition in presence severe occlusions. participants free use any training data testing built organisers synthetically occluding...
Presentation attacks are recurrent threats to biometric systems, where impostors attempt bypass these systems. Humans often use background information as contextual cues for their visual system. Yet, regarding face-based the is discarded, since face presentation attack detection (PAD) models mostly trained with crops. This work presents a comparative study of PAD (including multi-task learning, adversarial training and dynamic frame selection) in two settings: without The results show that...
With the ever-growing complexity of deep learning models for face recognition, it becomes hard to deploy these systems in real life. Researchers have two options: 1) use smaller models; 2) compress their current models. Since usage might lead concerning biases, compression gains relevance. However, compressing be also responsible an increase bias final model. We investigate overall performance, performance on each ethnicity subgroup and racial a State-of-the-Art quantization approach when...
The development of deep learning algorithms has extensively empowered humanity's task automatization capacity. However, the huge improvement in performance these models is highly correlated with their increasing level complexity, limiting usefulness human-oriented applications, which are usually deployed resource-constrained devices. This led to compression techniques that drastically reduce computational and memory costs without significant degradation. paper aims systematize current...
Manual assessment of fragments during the pro-cessing pathology specimens is critical to ensure that material available for slide analysis matches captured grossing without losing valuable this process. However, step still performed manually, resulting in lost time and delays making complete case evaluation by pathologist. To overcome limitation, we developed an autonomous system can detect count number contained on each slide. We applied compared two different methods: conventional machine...
Morphing attacks are one of the many threats that constantly affecting deep face recognition systems. It consists selecting two faces from different individuals and fusing them into a final image contains identity information both. In this work, we propose novel regularisation term takes account existent in both promotes creation orthogonal latent vectors. We evaluate our proposed method (OrthoMAD) five types morphing FRLL dataset performance model when trained on distinct datasets. With...