David Pascual-Hernández

I'm a computer vision engineer and researcher who likes solving hard perception problems and seeing them work in the real world. Over 6+ years I've designed and shipped production computer vision systems, including textura.ai's material digitization pipeline at SEDDI, where I led the R&D Optical Team, and contributed to research published at CVPR, SIGGRAPH, and other venues. My work spans CNNs, Vision Transformers, and GANs for segmentation, image restoration/enhancement, and object detection, with a recurring focus on evaluating model robustness and generalization across datasets and domains.

I'm currently pursuing a PhD at Rey Juan Carlos University on multimodal perception for autonomous navigation in off-road environments, and I enjoy moving fast from proof-of-concept to validated, deployable solutions.

Research Assistant at Rey Juan Carlos University (2024-10 – Present)

• Research on deep learning models, including Vision Transformers and CNNs, for real-time semantic segmentation using RGB, LiDAR, and multimodal sensor fusion, focused on robust perception and autonomous navigation in challenging off-road environments.
• Led a cross-dataset evaluation study benchmarking segmentation models' generalization and robustness across datasets and domains, proposing a unified ontology to enable fair comparison and reliable performance assessment (published in Expert Systems with Applications).
• Developed synthetic data generation pipelines in CARLA to create diverse virtual driving scenarios for training and evaluating perception models under controlled environmental conditions.
• Main maintainer of PerceptionMetrics, an open-source toolkit for evaluating object detection and segmentation models across sensor modalities, frameworks, and datasets.

Optical Team Lead at SEDDI (2023-01 – 2024-10)

• Led an R&D team of 5 computer vision/ML engineers while remaining hands on across the full machine learning lifecycle, from proof-of-concept and dataset acquisition to model development, deployment, and production validation.
• Developed and deployed deep learning models based on Vision Transformers and GANs for image restoration and image-to-image translation, balancing accuracy, inference latency, robustness, and scalability.
• Partnered with product management to translate business requirements into validated PoCs and deployable computer vision solutions, presenting experimentation and evaluation results to both technical and business stakeholders to guide deployment decisions.

Computer Vision Engineer at SEDDI (2018-08 – 2023-01)

• Contributed to the architecture, development, and scaling of textura.ai from its inception through production deployment, reaching multiple paying customers.
• Designed components of a multi-camera, multi-illumination gonioreflectometer for production-grade textile material digitization.
• Built robust color constancy and calibration pipelines for imaging systems operating under strict reliability and performance constraints.

Immersive Video Researcher Intern at Nokia Bell Labs (2017-10 – 2018-08)

• Developed hand gesture recognition interfaces for human-computer interaction.
• Researched augmented and mixed reality technologies for immersive video applications.

Machine Vision Engineer Intern at Robert Bosch (2016-08 – 2017-01)

• Evaluated industrial Data Matrix reading systems, imaging hardware, and optical lenses.
• Designed and validated machine vision inspection systems using NeuroCheck and custom optical design tools.

PhD in Artificial Vision, Pattern Recognition and Image Processing – Rey Juan Carlos University (2024)

MSc in Computer Vision – Rey Juan Carlos University (2017 – 2020)

BSc in Audiovisual and Multimedia Systems Engineering – Rey Juan Carlos University (2012 – 2017)

ERASMUS+ in Electrical Engineering – Óbuda University (2016 – 2016)