About Me

Hello! I am Nerea Gallego, a Computer Science PhD Candidate at the University of Zaragoza, currently working within the Robotics, Computer Vision and AI Group (RoPert) supported by a DGA Fellowship.

My research revolves around a central question: How can we help machines understand complex human behaviors using multiple sensors? Specifically, I specialize in multimodal learning, open-vocabulary Vision-Language Models (VLMs), and event-based vision. I am passionate about applying these cutting-edge AI technologies to solve real-world challenges, particularly in the medical field, such as improving patient monitoring in sleep studies through neuromorphic cameras to preserve privacy and work in low-light conditions.

Beyond the code and the lab, I am a strong advocate for making technology accessible. Whether I am co-supervising undergraduate theses, sharing my knowledge as a STEM tutor, or volunteering for the "Girls' Day" to inspire the next generation of women in engineering, I believe that the true power of AI lies in its community and its positive impact on society.

My current focus areas:

  • Multimodal Human Behavior Inference: Understanding social interactions and multi-agent dynamics in videos.
  • Neuromorphic Vision in Healthcare: Developing architectures like EventSleep2 for full-night clinical monitoring.
  • STEM Outreach: Bridging the gap between complex AI concepts and young minds.

Resume

Education

2024 — December 2028 (expected)
PhD in Computer Science - University of Zaragoza
My thesis is called "Understanding complex dynamic scenes with multiple sensors", and my advisor is Eduardo Montijano Muñoz. I work on deep learning for video and image understanding, vision-language models, temporal modeling, uncertainty estimation and event cameras.
2023 — 2025
Master in Robotics, Graphics and Computer Vision - University of Zaragoza
Master's Thesis: Semantic Recognition of Multi Agent Interactions in Videos.
2019 — 2023
Bachelor's Degree in Computer Science - University of Zaragoza
Bachelor's thesis: Patient monitoring in sleep studies with event cameras.

Experience

January 2025 — Present
Predoctoral Researcher (DGA Contract) — RoPert (UNIZAR)
I am building VLM frameworks for open-vocabulary human–object interaction detection, with first experiments on egocentric video (Ego4D). I built TIMID, which uses temporal constraints between actions to find mistakes in videos of robots (IROS 2026), and I am now applying the same idea to multi-target tracking. I also finished EventSleep2, which recognises activities over full nights of event-camera recordings and gives an uncertainty for each prediction (CVIU 2026). I am first author on both papers, and I do the PyTorch models, data pipelines, experiments and writing.
February 2023 — December 2024
Research Intern and Research Assistant — RoPert (UNIZAR - I3A)
I worked on human–object interaction detection with VLMs and studied where they fail zero-shot on unseen interactions, which became the start of my Master's thesis. I built a real-time prototype that tells users how to correct a sensor's placement (ROBOT 2024, first author), and I led the architecture of EventSleep: the event-camera pipeline, a new dataset and baselines. The code is public (ECCV Workshops 2024).
July 2022 - September 2022
Bitbrain (neurotechnology) — Research Intern
I implemented AI pipelines to detect slow waves in EEG signals, together with the neurotechnology experts at the company.

Grants and awards

  • PhD Scholarship (DGA)

    The predoctoral contract from the Government of Aragon (DGA) that pays for my PhD.

  • PhD Starting Scholarship (I3A)

    A scholarship from the Aragon Institute of Engineering Research (I3A) for starting my PhD.

  • Best Bachelor Thesis Award (NTT DATA)

    NTT DATA gave me this award for my Bachelor's thesis on monitoring patients in sleep studies with event cameras.

  • Bachelor Thesis Scholarship (I3A)

    An I3A scholarship for working on my Bachelor's thesis.

Languages

Spanish 100%
English (B2, Cambridge FCE; daily research use) 80%

Skills

  • Machine Learning

    PyTorch, deep neural networks for video and images, VLMs (CLIP, LLaVA), Hugging Face, uncertainty estimation, Sklearn, NumPy, Pandas.

  • Computer Vision

    Video and image processing, OpenCV, activity recognition, HOI, tracking, event cameras, real-time prototypes.

  • Software

    Python (advanced), C/C++, Java (object-oriented design), Matlab, SQL; Linux, shell scripting, Git, Docker.

Tech Stack & Expertise

Python PyTorch Computer Vision Machine Learning Event Cameras Robotics

Service, teaching and outreach

  • Reviewer

    I have reviewed papers for the journal Pattern Recognition and for two conferences, IROS 2025 and the IEEE International Conference on Robotics and Automation (ICRA) 2026.

  • Teaching

    I help teach the Artificial Intelligence course in the Computer Science degree at UNIZAR, and I have co-supervised two Bachelor's theses.

  • Invited talk (industry)

    I gave a training session to the R&D team at Mecanus Innovación S.L., in collaboration with Zebra Ventures S.L. It covered new technologies from research and how to apply them in industry, and my part was multi-sensor tracking and anomaly detection for industrial settings.

  • Schools

    I was selected for the ELIAS–ELLIS–VISMAC Winter School on multimodal learning.

  • Outreach

    I volunteer in the University of Zaragoza's Engineering Week (Girls' Day) and in Una ingeniera en cada cole, where I give robotics and AI demos to get children interested in STEM.

Publications and projects

TIMID: Time-Dependent Mistake Detection in Videos of Robot Executions

IROS 2026 | • 2026

Nerea Gallego*, Fernando Salanova*, Claudio Mannarano, Cristian Mahulea, Eduardo Montijano

TIMID Teaser

Sometimes a robot does every action of a task correctly but at the wrong time. TIMID detects these mistakes in videos of robot executions by using the temporal constraints between actions.

Paper Project Website

EventSleep2: Sleep activity recognition on complete night sleep recordings with an event camera

CVIU | • 2025

Nerea Gallego*, Carlos Plou*, Miguel Marcos, Pablo Urcola, Luis Montesano, Eduardo Montijano, Ruben Martinez-Cantin, Ana C. Murillo

EventSleep2 Teaser

This is the follow-up to EventSleep, now on recordings of complete nights instead of short clips. I handled the acquisition and processing of the full-night recordings. The model also gives an uncertainty estimate for each prediction, using Bayesian deep learning.

Paper Project Website

Vision-based feedback on correct sensor placement in medical studies

ROBOT | • 2024

Nerea Gallego, Carlos Plou, Luis Montesano, Ana C. Murillo, Eduardo Montijano

ROBOT 2024 Teaser

A real-time system that uses a camera to tell the user how to correct the placement of an EEG sensor. I calibrated the sensors, collected data from several subjects and designed the evaluation with users in the loop.

Paper Project Website

EventSleep: Sleep activity recognition with event cameras

ECCV Workshop | • 2024

Carlos Plou, Nerea Gallego, Alberto Sabater, Eduardo Montijano, Pablo Urcola, Luis Montesano, Ruben Martinez-Cantin, Ana C. Murillo

EventSleep Teaser

We recorded people sleeping with an event camera, which works in low light and is easier to use around patients' privacy than a normal camera. The paper has a new dataset, baselines for recognising sleep activities, and the code. I led the architecture.

Paper Project Website

Contact

Email is the best way to reach me. I am also on GitHub and LinkedIn.

Email: ngallego@unizar.es

GitHub: github.com/ngallego

LinkedIn: linkedin.com/in/nerea-gallego