Oihana Otaegui

• PhD in Telecommunication
• Head of Department at Vicomtech

In recent years, we have witnessed significant progress in emerging deep learning models, particularly Large Language Models (LLMs) and Vision-Language Models (VLMs). These models have demonstrated promising results, indicating a new era of Artificial Intelligence (AI) that surpasses previous methodologies. Their extensive knowledge and zero-shot c...

Gaze-annotated facial data is crucial for training deep neural networks (DNNs) for gaze estimation. However, obtaining these data is labor-intensive and requires specialized equipment due to the challenge of accurately annotating the gaze direction of a subject. In this work, we present a generative framework to create annotated gaze data by levera...

Gaze-annotated facial data is crucial for training deep neural networks (DNNs) for gaze estimation. However, obtaining these data is labor-intensive and requires specialized equipment due to the challenge of accurately annotating the gaze direction of a subject. In this work, we present a generative framework to create annotated gaze data by levera...

Deep neural network (DNN)-based vision systems could improve passenger transportation safety by automating processes such as verifying the correct positioning of luggage, seat occupancy, etc. Abundant and well-distributed data are essential to make DNNs learn appropriate pattern recognition features and have enough generalization ability. The use o...

Today's vehicles are increasingly embedded with computers and sensors which produce huge amount of data. The data are exploited for internal purposes and with the development of connected infrastructures and smart cities, the vehicles interact with each other as well as with road users generating other types of data. The access to these data and in...

Work on Local Dynamic Maps (LDM) implementation is still in its early stages, as the LDM standards only define how information shall be structured in databases, while the mechanism to fuse or link information across different layers is left undefined. A working LDM component, as a real-time database inside the vehicle is an attractive solution to m...

Strategies that include the generation of synthetic data are beginning to be viable as obtaining real data can be logistically complicated, very expensive or slow. Not only the capture of the data can lead to complications, but also its annotation. To achieve high-fidelity data for training intelligent systems, we have built a 3D scenario and set-u...

The detection and recognition of text instances in camera-captured images or videos generate rich and precise semantic information for interpreting and describing the scene. However, recognizing text in the wild remains a challenging problem despite its value. Apart from the inherent problems in text detection and recognition tasks, knowing the lan...

Cars capture and generate huge volumes of data in real-time about the driving dynamics, the environment, and the driver and passengers' activities. Due to the proliferation of cooperative, connected and automated mobility (CCAM), the value of data from vehicles is getting strategic, not just for the automotive industry, but also for many diverse st...

Tremendous advances in advanced driver assistance systems (ADAS) have been possible thanks to the emergence of deep neural networks (DNN) and Big Data (BD) technologies. Huge volumes of data can be managed and consumed as training material to create DNN models which feed functions such as lane keeping systems (LKS), automated emergency braking (AEB...

Local dynamic map (LDM) is a key component in the future of autonomous and connected vehicles. An LDM serves as a local database with the necessary tools to have a common reference system for both static data (i.e., map information) and dynamic data (vehicles, pedestrians, etc.). The LDM should have a common and well-defined input system in order t...

Driver Monitoring Systems (DMS) operate by measuring the state of the driver while performing driving activities. At the gates of the arrival of SAE-L3 autonomous driving vehicles, DMS are called to play a major role for guarantee or, at least, support safer mode transfer transitions (between manual and automated driving modes). Drowsiness and fati...

Modern Artificial Intelligence (AI) methods can produce a large quantity of accurate and richly described data, in domains such as surveillance or automation. As a result, the need to organize data at a large scale in a semantic structure has arisen for long-term data maintenance and consumption. Ontologies and graph databases have gained popularit...

To train Deep Neural Networks (DNNs)-based methods, suitable training data are key to help DNNs learn appropriate pattern recognition features. The use of synthetic data may help in generating sufficient and balanced data. However, models trained with such data often present a domain gap when applied to real-world scenarios. Many studies focus on t...

We present an approach to optimally deploy Deep Neural Networks (DNNs) in serverless cloud architectures. A serverless architecture allows running code in response to events, automatically managing the required computing resources. However, these resources have limitations in terms of execution environment (CPU only), cold starts, space, scalabilit...

Current 3D object detectors from Bird’s Eye View (BEV) LiDAR point cloud data rely on Convolutional Neural Networks (CNNs), which have originally been designed for camera images. Therefore, they look for the same target features, regardless of the position of the objects with respect to the sensor. Discarding this spatial information makes 3D objec...

Data labeling has become a major problem in industries aiming to create and use ground truth labels from massive multi-sensor archives to feed into Artificial Intelligence (AI) applications. Annotation of multi-sensor set-ups with multiple cameras and LIDAR is now particularly relevant for the automotive industry aiming to build Autonomous Driving...

The recently presented Driver Monitoring Dataset (DMD) extends research lines for Driver Monitoring Systems. We intend to explore this dataset and apply commonly used methods for action recognition to this specific context, from image-based to video-based analysis. Specially, we aim to detect driver distraction by applying action recognition techni...

Lane markings are a key element for Autonomous Driving. The generation of high definition maps and ground-truth data require extensive manual labor. In this paper, we present an efficient and robust method for the offline annotation of lane markings, using low-density LIDAR point clouds and odometry information. The odometry is used to accumulate t...

Current computing requirements for high-scale inference of Deep Neural Networks (DNNs) demand distributed execution environments. The advantages of serverless functions in distributed computation offloading and automatic resource scalability make them a very suitable environment for such a task. However, finding the optimal workload at minimum reso...

Vision is the richest and most cost-effective technology for Driver Monitoring Systems (DMS), especially after the recent success of Deep Learning (DL) methods. The lack of sufficiently large and comprehensive datasets is currently a bottleneck for the progress of DMS development, crucial for the transition of automated driving from SAE Level-2 to...

An innovative solution named Annotation as a Service (AaaS) has been specifically designed to integrate heterogeneous video annotation workflows into containers and take advantage of a cloud native highly scalable and reliable design based on Kubernetes workloads. Using the AaaS as a foundation, the execution of automatic video annotation workflows...

The lane-level localization of vehicles with low-cost sensors is a challenging task. In situations in which Global Navigation Satellite Systems (GNSSs) suffer from weak observation geometry or from the influence of reflected signals, the fusion of heterogeneous information presents a suitable approach for improving the localization accuracy. We pro...

Face fitting methods align deformable models to faces on images using the information given by the image pixels. However, most algorithms are designed to be used in desktop personal computers (PC), or hardware with significant computational power. These approaches are therefore too demanding for devices with limited computational power, like the in...

Tracking urban mobility with current heterogeneous sensing capabilities has opened a wide research area on analytical and predictive data-driven models for improvements in transport operations and planning. These improvements are applicable for individual users, service providers and decision-makers. People, vehicles and goods move along the city a...

This paper presents a new approach to 3D object detection that leverages the properties of the data obtained by a LiDAR sensor. State-of-the-art detectors use neural network architectures based on assumptions valid for camera images. However, point clouds obtained from LiDAR are fundamentally different. Most detectors use shared filter kernels to e...

Vision is the richest and most cost-effective technology for Driver Monitoring Systems (DMS), especially after the recent success of Deep Learning (DL) methods. The lack of sufficiently large and comprehensive datasets is currently a bottleneck for the progress of DMS development, crucial for the transition of automated driving from SAE Level-2 to...

Current real-time data collection systems for urban transportation and mobility allow enhancing digital maps with up-to-date situational information. This information is of great interest for short-term navigation and route planning as well as for medium-to long-term mobility data analysis , as it provides a finer time-varying detail of the urban m...

This paper introduces a web application for point cloud annotation that is used in the advanced driver assistance systems field. Apart from the point cloud viewer, the web tool has an object viewer and a timeline to define the attributes of the annotations and a video viewer to validate the point cloud annotations with the corresponding video image...

Significant efforts have been made and are still being made on short-term traffic prediction methods, specially for highway traffic based on punctual measurements. Literature on predicting the spatial distribution of the traffic in urban intersections is, however, very limited. This work presents a novel data-driven prediction algorithm based on Ra...

GNSS, for many very good reasons among which continuity, is hybridized for land positioning. Indeed, GNSS outages remain frequent in the core of modern cities, despite satellite multi-constellation inter-operating today. Urban positioning is known as a difficult problem, in which satellite measurements, if not occulted, suffer from multipath caused...

Multimodal deep learning is about learning features over multiple modalities. Impressive progress has been made in deep learning solutions that rely on a single sensor modality for advanced driving. However, these approaches are limited to cover certain functionalities. The potential of multimodal sensor fusion has been very little exploited, altho...

Despite the increasing hardware capabilities of embedded devices, running a Deep Neural Network (DNN) in such systems remains a challenge. As the trend in DNNs is to design more complex architectures, the computation time in low-resource devices increases dramatically due to their low memory capabilities. Moreover, the physical memory used to store...

Novel ubiquitous traffic sensors such as floating car data (FCD) are getting extended due to the use of 24 h connected smartphones and global positioning systems. Road conditions such as travel speeds in each road link and mobility demand can be monitored by measurements coming from moving vehicles consisting of geolocation and speed information wi...

Automated driving will have a big impact on society, creating new possibilities for mobility and reducing road accidents. Current developments aim to provide driver assistance in the form of conditional and partial automation. Computer Vision, either alone or combined with other technologies such as radar or Lidar, is one of the key technologies of...

Current 4G networks are approaching the limits of what is possible with this generation of radio technology. Future 5G networks will be highly based on software, with the ultimate goal of being self-managed. Machine Learning is a key technology to reach the vision of a 5G self-managing network. This new paradigm will significantly impact on connect...

Technology roadmapping provides a strategic tool to help companies develop an outside-in view and challenge their current competitive perspectives. In this paper, the authors describe the roadmapping process, which is aligned, with the research and development (R&D) strategy of an applied research centre. This process is based in an adapted combina...

Computer vision methods for advanced driver assistance systems (ADAS) must be developed considering the strong requirements imposed by the industry, including real-time performance in low cost and low consumption hardware (HW), and rapid time to market. These two apparently contradictory requirements create the necessity of adopting careful develop...

Computer Vision, either alone or combined with other technologies such as radar or Lidar, is one of the key technologies used in Advanced Driver Assistance Systems (ADAS). Its role understanding and analysing the driving scene is of great importance as it can be noted by the number of ADAS applications that use this technology. However, porting a v...

In this study, the authors analyse the exponential growth of advanced driver assistance systems based on video processing in the past decade. Specifically, they focus on how research and innovative ideas can finally reach the market as cost-effective solutions. They explore well-known computer vision methods for services like lane departure warning...

There are many reasons for programming in Tizen: it is based on standards, is open, it has industry support and multiple profiles. In this paper we test the potential of one this profiles: Tizen IVI (In-Vehicle Infotainment). For this purpose, we present the implementation of two Computer Vision based Advanced Driver Assistance Systems (ADAS) in a...

Embedding computer vision SW for in-vehicle applications requires the optimization of algorithms to fit into low cost and low consumption HW. Such optimization is a task substantially centered in improving the efficiency of the implementations, typically focused on the migration of algorithms to massive parallelization HW. The development cost asso...

This work addresses the challenge of designing an effective, reliable and affordable autonomous navigation system for blind and visually impaired people which also covers journey planning and post journey activities (such as recommendations and experiences sharing) . The main contribution focuses on the integration of accurate real-time user positi...

Bird migration behaviours and patterns offer invaluable information about environmental evolution of regions. Traditional bird ringing techniques are often used for nature monitoring. A web based geographical visualization and analysis tool is presented in this paper, showing how cartographic data of wetland areas, their vegetation and water facili...

Most recent visual odometry algorithms based on sparse feature matching are computationally efficient methods that can be executed in real time on desktop computers. However, further efforts are required to reduce computational complexity in order to integrate these solutions in embedded platforms with low power consumption. This paper presents a s...

Computer vision technologies can contribute in many ways to the development of smart cities. In the case of vision applications for advanced driver assistance systems (ADAS), they can help to increase road traffic safety, which is a major concern nowadays. The design of an embedded vision system for driver assistance is not straightforward; several...

In this paper we analyse the recent exponential growth of applications based on video processing in the framework of Advanced Driver Assistance Systems (ADAS). Specifically, we focus on how such cost-effective solutions can finally reach the market starting from research and innovative ideas. We explore well known computer vision methods for servic...

In this paper we analyze the recent exponential growth of applications based on video processing in the framework of Advanced Driver Assistance Systems (ADAS). Specifically, we focus on the cost-effective solutions provided by computer vision methods for services like lane departure warning systems, collision avoidance systems, etc. Along the paper...

An automatic method for rail inspection is introduced in this paper. The method detects rail flaws using computer vision algorithms. Unlike other methods designed for the same goal, we propose a method that automatically fits a 3D rail model to the observations. The proposed strategy is based on the novel combination of a simple but effective laser...

People with visual impairments have striking needs for trustful navigation systems enabling for efficient mobility services, mainly considering safely and autonomy. In this context, satellite-positioning and navigation technologies are being implemented leading to innovative personal navigation devices. Existing products and solutions based on GNSS...

Although automatic and driverless vehicles do operate nowadays, most of these systems are currently guided by railwaysor magnetic wires. More recently, a few demonstrations have been undertaken in order to demonstrate the capabilities of satellite based navigation systems to clear the roads from such heavy infrastructures. TAXISAT aims at developin...

This work addresses the challenge of designing an effective, reliable and affordable navigation system for blind and visually impaired people (BVIP). Our contribution focuses essentially on the integration of accurate real-time user positioning data with binaural 3D audio based guiding techniques on mobile devices. The purpose is to produce a binau...

A novel approach to real-time lane modeling using a single camera is proposed. The proposed method is based on an efficient design and implementation of a particle filter which applies the concepts of the Rao-Blackwellized particle filter (RBPF) by separating the state into linear and non-linear parts. As a result the dimensionality of the problem...

This paper introduces a rail inspection system which detects rail flaws using computer vision algorithms. Unlike other methods designed for the same purpose, we propose a method that automatically fits a 3D rail model to the observations during regular services and normal traffic conditions. The proposed strategy is based on a novel application of...

This article introduces a 3D vehicle tracking system in a traffic surveillance environment devised for shadow tolling applications. It has been specially designed to operate in real time with high correct detection and classification rates. The system is capable of providing accurate and robust results in challenging road scenarios, with rain, traf...

On-demand public transport satisfies many of the needs that traditional public transport, with fixed predeter- mined routes and schedules, cannot fulfill. Both schemas are able to work together, as it is discussed in this paper, being on- demand services a good alternative in the areas and time inter- vals where low demand makes fixed routes unsust...

Fires and other related disasters provoke great destruction of high valuable environments and economical losses, especially when they are located in urban areas. In this work, we present a combined urban and forest fire spreading algorithm to be used in real time and interactive virtual simulations. The algorithm is pedagogical oriented and its pur...

A new method for 3D vehicle modeling in low-cost monocamera surveillance systems is introduced in this paper. The proposed algorithm aims to resolve the projective ambiguity of 2D image observations by means of the integration of temporal information and model priors within a Markov Chain Monte Carlo (MCMC) method. The method is specially designed...

Due to the high complexity of the required calculations, Intelligent Routing Systems have to apply latest Operations Research techniques to be able to create routes efficiently. This paper proposes a solution to the Multi Path Orienteering Problem with Time Windows (MPOPTW), which includes multiple paths to move between locations. The main characte...