As road transportation has been identified as a major contributor of environmental pollution, mot... more As road transportation has been identified as a major contributor of environmental pollution, motivating individuals to adopt a more eco-friendly driving style could have a substantial ecological as well as financial benefit. With gamification being an effective tool towards guiding targeted behavioural changes, the development of realistic frameworks delivering a high end user experience, becomes a topic of active research. This paper presents a series of enhancements introduced to an eco-driving gamification platform by the integration of additional wearable and vehicle-oriented sensing data sources, leading to a much more realistic evaluation of the context of a driving session.
The deployment of high performing deep learning models on platforms of limited resources is curre... more The deployment of high performing deep learning models on platforms of limited resources is currently an active area of research. Among the main directions followed so far, pre-trained neural networks are accelerated and compressed by appropriately modifying their structure and / or parameters. Capitalizing on a recently proposed codebook of a special structure that can be utilized in the frame of the so-called weight sharing methods, this paper describes a "data-driven" technique for designing such a codebook. The performance of the technique, in terms of the observed representation error and classification accuracy versus the achieved acceleration ratio, is demonstrated by considering the VGG16 and the ResNet18 models, pre-trained on the ILSVRC2012 dataset.
Zenodo (CERN European Organization for Nuclear Research), May 29, 2020
Recent advances in 3D scanning technology have enabled the deployment of 3D models in various ind... more Recent advances in 3D scanning technology have enabled the deployment of 3D models in various industrial applications like digital twins, remote inspection and reverse engineering. Despite their evolving performance, 3D scanners, still introduce noise and artifacts in the acquired dense models. In this work, we propose a fast and robust denoising method for dense 3D scanned industrial models. The proposed approach employs conditional variational autoencoders to effectively filter face normals. Training and inference are performed in a sliding patch setup reducing the size of the required training data and execution times. We conducted extensive evaluation studies using 3D scanned and CAD models. The results verify plausible denoising outcomes, demonstrating similar or higher reconstruction accuracy, compared to other state-of-the-art approaches. Specifically, for 3D models with more than 1e4 faces, the presented pipeline is twice as fast as methods with equivalent reconstruction error.
Zenodo (CERN European Organization for Nuclear Research), Jul 6, 2020
Throughout the years, several works have been proposed for 3D mesh denoising. Nevertheless, despi... more Throughout the years, several works have been proposed for 3D mesh denoising. Nevertheless, despite their reconstruction quality, there are still challenges related to the preservation of the fine surface features. Motivated by the impressive results of image denoising by 3D transform-domain collaborative filtering (CF), we extend it to 3D mesh denoising. CF is also capable of revealing the finest details shared by grouped blocks while preserving at the same time the unique features of each block. A new promising approach suggests unrolling the computational pipeline of CF into a convolutional neural network (CNN) structure increasing significantly the efficiency of this solution. In this paper, we successfully extend and apply this method to 3D meshes making a transition from face normals to pixels. Extensive evaluation studies carried out using a variety of 3D meshes verify that the proposed approach achieves plausible reconstruction outputs and provides very promising results.
Zenodo (CERN European Organization for Nuclear Research), Jul 8, 2021
Cooperative autonomous driving in 5G and smart cities environment is expected to further improve ... more Cooperative autonomous driving in 5G and smart cities environment is expected to further improve safety, security and efficiency of transportation systems. To this end, involved vehicles is imperative to have accurate knowledge of both their own and neighboring vehicles' location, a task known as cooperative awareness. In this paper, we have formulated two novel distributed localization and tracking schemes, based on Gradient Descent and Extended Kalman Filter algorithms, to cope with erroneous GPS location. Sensor-rich vehicles exploit Vehicle-to-Vehicle communications and a multitude of integrated sensors, like LIDAR and Cameras, to generate and fuse heterogeneous data. Each vehicle interacts only with its own connected neighboring vehicles, formulating individual star topologies. Extensive simulation studies using CARLA autonomous driving simulator, verify the significant reduction of GPS error achieved by the two methods in various experimental conditions. Distributed tracking proves to be much superior than Gradient descent algorithm, both in the case of self (58% reduction of GPS error) and neighboring vehicles location estimation (38% reduction of average GPS error).
Fully autonomous vehicles may still be an elusive goal, however, research in the deployment of re... more Fully autonomous vehicles may still be an elusive goal, however, research in the deployment of relevant Artificial Intelligence technologies in the domain is rapidly gaining traction. A key challenge lies in the fusion of all the diverse information from the various sensors on the vehicle and its environment. In this context, ontologies and semantic technologies can effectively address this challenge by semantically fusing heterogeneous pieces of information into a uniform Knowledge Graph. This paper presents CASPAR, an extensible semantic data fusion platform for autonomous vehicles. Two use case scenarios are also presented that demonstrate the framework's versatility.
Through the years, several works have demonstrated high-quality 3D mesh denoising. Despite the hi... more Through the years, several works have demonstrated high-quality 3D mesh denoising. Despite the high reconstruction quality, there are still challenges that need to be addressed ranging from variations in configuration parameters to high computational complexity. These drawbacks are crucial especially if the reconstructed models have to be used for quality check, inspection or repair in manufacturing environments where we have to deal with large objects resulting in very dense 3D meshes. Recently, deep learning techniques have shown that are able to automatically learn and find more accurate and reliable results, without the need for setting manually parameters. In this work, motivated by the aforementioned requirements, we propose a fast and reliable denoising method that can be effectively applied for reconstructing very dense noisy 3D models. The proposed method applies conditional variational autoencoders on face normals. Extensive evaluation studies carried out using a variety of 3D models verify that the proposed approach achieves plausible reconstruction outputs, very relative or even better of those proposed by the literature, in considerably faster execution times.
Cooperative Localization is expected to play a crucial role in various applications in the field ... more Cooperative Localization is expected to play a crucial role in various applications in the field of Connected and Autonomous vehicles (CAVs). Future 5G wireless systems are expected to enable cost-effective Vehicle-to-Everything (V2X) systems, allowing CAVs to share with the other entities of the network the data they collect and measure. Typical measurement models usually deployed for this problem, are absolute position from Global Positioning System (GPS), relative distance and azimuth angle to neighbouring vehicles, extracted from Light Detection and Ranging (LIDAR) or Radio Detection and Ranging (RADAR) sensors. In this paper, we provide a cooperative localization approach that performs multi modal-fusion between the interconnected vehicles, by representing a fleet of connected cars as an undirected graph, encoding each vehicle position relative to its neighbouring vehicles. This method is based on: i) the Laplacian Processing, a Graph Signal Processing tool that allows to capture intrinsic geometry of the undirected graph of vehicles rather than their absolute position on global coordinate system and ii) the temporal coherence due to motion patterns of the moving vehicles.
Zenodo (CERN European Organization for Nuclear Research), Aug 31, 2015
Wheezes are abnormal continuous adventitious lung sounds that are strongly related to patients wi... more Wheezes are abnormal continuous adventitious lung sounds that are strongly related to patients with obstructive airways diseases. Wireless telemonitoring of these sounds facilitate early diagnosis (short, long term) and management of chronic inflammatory disease of the airways (e.g., asthma) through the use of an accurate and energy efficient mhealth system. Therefore, low complexity breath compression schemes with high compression ratio are required. To this end, we propose a compressed sensing based compression/reconstruction solution that enables wheeze detection from a small number of linearly encoded samples, by exploiting the block sparsity of the breath eigenspectrum during reconstruction at the receiver. Simulation studies, carried out with publicly available breath sounds, show the energy efficiency benefits of the proposed CS scheme, compared to traditional CS recovery approaches.
Road transport is one of the major causes of the environmental pollution. Among the actions indiv... more Road transport is one of the major causes of the environmental pollution. Among the actions individuals can take to reduce their green-house gases associated with personal transportation, there is to operate their current vehicles more efficiently. Behavioral theory strongly confirms that the most important educational element in changing driver behavior is the direct feedback while driving on an immediate and continuous basis. Gamification has been positioned as a powerful approach, tool, or set of techniques that guides targeted behavior change and improves the way that various activities are undertaken so that those involved begin to take the desired actions while they experience more fun, enjoyment, and pleasure in their tasks. Building on this direction, we present conceptual approach of an eco-driving simulation system that aims to train drivers to follow eco-driving rules simulating the augmented reality technology in virtual driving games. The proposed system provides: i) an efficient way to study the effect of AR games responsible for monitoring driving behavior and delivering action personalized plans that will help user to maintain a green driving style without distracting them from safe driving and ii) a multiplayer gaming environment where users can monitor the eco-driving score evolution, set missions and invite other to participate collaboratively or competitively.
Outliers Removal of Highly Dense and Unorganized Point Clouds Acquired by Laser Scanners in Urban Environments
Recently, there is a tremendous interest in the processing of unorganized point clouds, generated... more Recently, there is a tremendous interest in the processing of unorganized point clouds, generated using a variety of 3D scanning technologies such as structured light and LIDAR systems. Without a doubt, the most compelling problem in this domain is the removal of outliers. To effectively address the aforementioned issue, we present a novel method, that detects accurately and efficiently the outliers by exploiting the spatial coherence in the object geometry and the sparsity of the outliers in the spatial domain. This is achieved by solving a convenient convex method called Robust PCA (RPCA). To demonstrate the effectiveness of the proposed technique, we evaluate it by using real scanned point clouds which are extremely dense consisting of millions of points.
Cooperative Localization has received extensive interest from several scientific communities incl... more Cooperative Localization has received extensive interest from several scientific communities including Robotics, Optimization, Signal Processing and Wireless Communications. It is expected to become a major aspect for a number of crucial applications in the field of Connected and (Semi-) Autonomous vehicles (CAVs), such as collision avoidance/warning, cooperative adaptive cruise control, safely navigation, etc. 5G mobile networks will be the key to providing connectivity for vehicle to everything (V2X) communications, allowing CAVs to share with other entities of the network the data they collect and measure. Typical measurement models usually deployed for this problem, are absolute position information from Global Positioning System (GPS), relative distance to neighbouring vehicles and relative angle or azimuth angle, from Light Detection and Ranging (LIDAR) or Radio Detection and Ranging (RADAR) sensors. In this paper, we provide a cooperative estimation approach that performs multi modal-fusion between interconnected vehicles. This method is based on a Graph Signal Processing tool, known as Laplacian Graph Processing, and significantly outperforms existing method both in terms of attained accuracy and computational complexity.
3D Mesh Inpainting Using Matrix Completion via Augmented Lagrange Multiplier Method
Recently, 3D objects have started to attract a lot of attention, due to their increased usage in ... more Recently, 3D objects have started to attract a lot of attention, due to their increased usage in a large variety of different fields (e.g., archeology, gaming, topography, medicine etc.), Despite the technological improvement that has been occurred, there are still limitations that can cause noisy or incomplete data. In this paper, we focus on the problems of partial-deformed or partial-observed 3D meshes, presenting a very fast and efficient method for surface inpainting of dense 3D meshes. In order to achieve this, we use a matrix completion approach taking advantage of the low-rank property attributed to spatial coherence. Two different scenarios are studied and presented: (i) the repairing of partial-observed 3D meshes consisting of incomplete surfaces, (ii) the recovery of 3D objects with holes which are created by the intentional removal of deformed areas. We demonstrate the performance of our approach providing experimental results for both cases highlighting the effectiveness of our method.
Spectral methods are widely used in geometry processing of 3D models. They rely on the projection... more Spectral methods are widely used in geometry processing of 3D models. They rely on the projection of the mesh geometry on the basis defined by the eigenvectors of the graph Laplacian operator, becoming computationally prohibitive as the density of the models increases. In this paper, we propose a novel approach for supporting fast and efficient spectral processing of dense 3D meshes, ideally suited for real time compression and denoising scenarios. To achieve that, we apply the problem of tracking graph Laplacian eigenspaces via orthogonal iterations, exploiting potential spectral coherences between adjacent parts. To avoid perceptual distortions when a fixed number of eigenvectors is used for all the individual parts, we propose a flexible solution that automatically identifies the optimal subspace size for satisfying a given reconstruction quality constraint. Extensive simulations carried out with different 3D meshes in compression and denoising setups, showed that the proposed schemes are very fast alternatives of SVD based spectral processing while achieving at the same time similar or even better reconstruction quality. More importantly, the proposed approach can be employed by several other state of the art denoising methods as a preprocessing step, optimizing both their reconstruction quality and their computational complexity.
With the growing demand for easy and reliable generation of 3D models representing real-world obj... more With the growing demand for easy and reliable generation of 3D models representing real-world objects and environments in mobile cloud computing platforms, new schemes for acquisition, storage and transmission of 3D meshes are required. In general, 3D meshes consist of two distinct components: vertex positions and vertex connectivity. Vertex position encoders are much more resource demanding than connectivity encoders, stressing the need for novel geometry compression schemes. The design of an accurate and energy efficient geometry compression system can be achieved by: i) reducing the amount of data that should be transmitted ii) minimizing the computational operations executed at the encoder. In this paper, we propose a Bayesian learning approach that allows processing large meshes in parts and reconstructing the Cartesian coordinates of each part from a small number of random linear combinations. The proposed compression/reconstruction approaches minimize the samples that are required for transmission yet assuring accurate reconstruction at the receiver, by exploiting specific local characteristics of the surface geometry in the graph Fourier domain. Simulation studies show that the proposed schemes, as compared to the state of the art approaches, achieve competitive Compression Ratios (CRs), offering at the same time significantly lower compression computational complexity, which is essential for mobile cloud computing platforms.
Zenodo (CERN European Organization for Nuclear Research), Sep 14, 2021
In this paper, we design distributed multi-modal localization approaches for Connected and Automa... more In this paper, we design distributed multi-modal localization approaches for Connected and Automated vehicles. We utilize information diffusion on graphs formed by moving vehicles, based on Adapt-then-Combine strategies coupled with the Least-Mean-Squares and the Conjugate Gradient algorithms. We treat the vehicular network as an undirected graph, where vehicles communicate with each other by means of Vehicle-to-Vehicle communication protocols. Connected vehicles perform cooperative fusion of different measurement modalities, including location and range measurements, in order to estimate both their positions and the positions of all other networked vehicles, by interacting only with their local neighborhood. The trajectories of vehicles were generated either by a well-known kinematic model, or by using the CARLA autonomous driving simulator. The proposed distributed and diffusion localization schemes significantly reduced the GPS error and do not only converged to the global solution, but they even outperformed it. Extensive simulation studies highlight the benefits of the various methods, which in turn outperform other state of the art approaches. The impact of the network connections and the network latency are also investigated.
Uploads
Papers by Aris Lalos