A5052453150- Smart Agriculture and AI
- Spectroscopy and Chemometric Analyses
- Remote Sensing in Agriculture
- Fetal and Pediatric Neurological Disorders
- Face recognition and analysis
- Advanced Vision and Imaging
- Ultrasound in Clinical Applications
- Human Pose and Action Recognition
- Video Surveillance and Tracking Methods
- Congenital Diaphragmatic Hernia Studies
- Domain Adaptation and Few-Shot Learning
- Neonatal Respiratory Health Research
- Preterm Birth and Chorioamnionitis
- Image and Video Stabilization
- Prenatal Screening and Diagnostics
- Anomaly Detection Techniques and Applications
- Face and Expression Recognition
- COVID-19 diagnosis using AI
- Leaf Properties and Growth Measurement
- Dental Radiography and Imaging
- Pregnancy and preeclampsia studies
- COVID-19 Clinical Research Studies
- Assisted Reproductive Technology and Twin Pregnancy
- Ectopic Pregnancy Diagnosis and Management
- Neonatal and fetal brain pathology
Universitat Oberta de Catalunya
2021-2025
Institute for Bioengineering of Catalonia
2024
Hospital Sant Joan de Déu Barcelona
2019-2023
Universitat de Barcelona
2019-2023
Hospital Clínic de Barcelona
2019-2022
California Institute of Technology
2011-2015
Technicolor (France)
2014-2015
Centre for Automation and Robotics
2007-2011
Consejo Superior de Investigaciones Científicas
2007-2011
Unidades Centrales Científico-Técnicas
2008
Human faces captured in real-world conditions present large variations shape and occlusions due to differences pose, expression, use of accessories such as sunglasses hats interactions with objects (e.g. food). Current face landmark estimation approaches struggle under since they fail provide a principled way handling outliers. We propose novel method, called Robust Cascaded Pose Regression (RCPR) which reduces exposure outliers by detecting explicitly using robust shape-indexed features....
A lack of automated, quantitative, and accurate assessment social behaviors in mammalian animal models has limited progress toward understanding mechanisms underlying interactions their disorders such as autism. Here we present a new integrated hardware software system that combines video tracking, depth sensing, machine learning for automatic detection quantification involving close dynamic between two mice different coat colors home cage. We designed setup integrates traditional cameras...
The goal of this study was to evaluate the maturity current Deep Learning classification techniques for their application in a real maternal-fetal clinical environment. A large dataset routinely acquired screening ultrasound images (which will be made publicly available) collected from two different hospitals by several operators and machines. All were manually labeled an expert maternal fetal clinician. Images divided into 6 classes: four most widely used anatomical planes (Abdomen, Brain,...
We present a novel method for analyzing social behavior. Continuous videos are segmented into action `bouts' by building temporal context model that combines features from spatio-temporal energy and agent trajectories. The is tested on an unprecedented dataset of interacting pairs mice, which was collected as part state-of-the-art neurophysiological study comprises over 88 hours (8 million frames) annotated videos. find our trajectory features, used in discriminative framework, more...
Abstract The objective of this study was to evaluate the performance a new version quantusFLM®, software tool for prediction neonatal respiratory morbidity (NRM) by ultrasound, which incorporates fully automated fetal lung delineation based on Deep Learning techniques. A set 790 ultrasound images obtained at 24 + 0–38 6 weeks’ gestation evaluated. Perinatal outcomes and occurrence NRM were recorded. quantusFLM® 3.0 applied all automatically delineate predict risk. test compared with same...
Only 30% of embryos from in vitro fertilized oocytes successfully implant and develop to term, leading repeated transfer cycles. To reduce time-to-pregnancy stress for patients, there is a need diagnostic tool better select based on their physiology. The current standard employs brightfield imaging, which provides limited physiological information. Here, we introduce METAPHOR: Metabolic Evaluation through Phasor-based Hyperspectral Imaging Organelle Recognition. This non-invasive, label-free...
This paper presents a mapping method for wide row crop fields. The resulting map shows the rows and weeds present in inter-row spacing. Because field videos are acquired with camera mounted on top of an agricultural vehicle, image sequence stabilization was needed consequently designed developed. proposed uses centers some as features to be tracked, which compensates lateral movement (sway) leaves pitch unchanged. A region interest is selected using tracked features, inverse perspective...
Generative adversarial networks (GANs) have been recently applied to medical imaging on different modalities (MRI, CT, X-ray, etc). However there are not many applications ultrasound modality as a data augmentation technique downstream classification tasks. This study aims explore and evaluate the generation of synthetic fetal brain images via GANs apply them improve plane classification. State art stylegan2-ada were image GAN-based classifiers compared with baseline classifiers. Our...
Numerous 'non-maximum suppression' (NMS) post-processing schemes have been proposed for merging multiple independent object detections.We propose a generalization of NMS beyond bounding boxes to merge pose estimates in single frame.The final are centroids rather than medoids as standard NMS, thus being more accurate any the individual candidates.Using same mathematical framework, we extend our approach multi-frame setting, across space and time outputting both number objects present...
One of the main issues current Head-Mounted Displays (HMD) is that they hide completely wearer's face. This can be an issue in social experiences where two or more users want to share 3D immersive experience. We propose a novel method recover face user real-time. First, we learn appearance offline by building textured model his head from series pictures. Then, calibrating camera and tracking HMD's position real-time reproject on top video frames mimicking exactly user's pose. Finally, remove...
Determination of the soil coverage by crop residues after ploughing is a fundamental element Conservation Agriculture. This paper presents application genetic algorithms employed during fine tuning segmentation process digital image with aim automatically quantifying residue coverage. In other words, objective to achieve that would permit discrimination texture so output binary in which zones are isolated from rest. The RGB images used come sample sections terrain were photographed...
Video-based activity and behavior analysis of mice has garnered wide attention in biomedical research. Animal facilities hold large numbers housed “home-cages” densely stored within ventilated racks. Automated their home-cages can provide a new set sensitive measures for detecting abnormalities time-resolved deviation from the baseline behavior. Large-scale monitoring animal requires minimal footprint hardware that integrates seamlessly with The compactness imposes use fisheye lenses...
This paper describes a new automatic image segmentation strategy for segmenting green plants. The final goal is its application in Precision Agriculture. to identify several classes of greenness coming from the We exploit performance existing approaches so that conveniently combined allow us design approach based on non methods. First we apply well known index-based accentuates spectral band remainder, giving gray image. From resulting well-known thresholding Otsu's method obtaining binary...