Tutorial on "Cooperative Smart Sensor Fusion and Differentiable Programming for Edge Devices"
Online from the University of Ghent (Belgium), December 14-16, 2020
As part of the training program n complementary technical skills, a tutorial was included in the program of WASC 2020. This tutorial covered the different aspects of fusion of information from sensors such as radars, cameras and LiDARs, what requires complicated analytics. The standard engineering approach to handle complexity is modularity: each “smart” sensor operates as a black box and the high level “decisions” or other outputs are fused in a separate “late fusion” module. However, this approach is sub-optimal. Early fusion combines only partially processed sensor data, rather than their final “decisions.” While provably optimal in terms of detecting, this approach has a major downside: it requires a lot more processing at the fusion centre and much more data communication from the sensors to the fusion centre. Cooperative fusion is a compromise between early and late fusion. It proposes loosely coupled sensor and inference modules that communicate over low bandwidth, low latency links. As in late fusion, the “smart” sensors take care of most of the processing tasks and produce heavily condensed outputs, e.g. candidate decisions and associated probabilities. However, the sensors now also receive auxiliary inputs from other sensors and/or from the fusion centre. This allows sensors to verify weak evidence by other sensors and tune their internal algorithms. The instructors of the course were:
- Wilfried Philips (imec-Ghent University).
- Bart Goosens (imec-Ghent University).
A total of 27 people attended the tutorial.
Workshop on the Architecture of Smart Cameras
Online from the University of Ghent (Belgium), December 14-16, 2020
The 9th edition of the Workshop on Architectures of Smart Cameras (WASC 2020) has been held by IMEC and Ghent University. Due to COVID19 restrictions, this edition will be an online event, using Zoom. The organizers of the event have been: Bart Goossens, Ljiljana Platisa, Wilfried Philips from imec-Ghent University; and Ricardo Carmona-Galán, from CSIC.
This year’s program covered:
- Vision sensors design and hardware
- Edge AI and cooperative camera systems
- Deep learning and visual inference
- Sensor-related image processing
It included also a session for interaction between different EU Projects that counted with the collaboration of coordinators and partners of projects: NextPerception, MENELAOS_NT, Comp4Drones and ACHIEVE-ITN.
It was attended by 27 people. All ESRs presented their latest results. Also researchers from Univ. Santiago de Compostela, Univ. Politécnica de Cartagena, INSA-Rennes, Univ. Clermont-Auvergne and Univ. Seville presented their works.
Career Guidance Panel Sessions
Online from the University of Seville, November 19 and 26, 2020
In parallel with scientific training, the PhD Program on Physical Sciences and Technologies (PDCyTF) of the University of Seville offers some career development guidance. This year it was organized by ACHIEVE-ITN’s Project Coordinator, and around 10 people involved in the Project participated in this activity. The activity was divided in two Career Guidance Panel Sessions:
Panel Session I - Technology transfer: from research results to products
Different aspects on technology transference were reviewed. The objective was to consider the transference of research results as a feasible path for professional development beyond PhD. Experiences on the development of innovative products/services/processes and the creation of technology-based start-ups were described and discussed.
- Juan Martínez Armesto (VATC-CSIC)
- Laura M. Lechuga (SENSIA S.L. and BIOD S.L.)
- Cruz Medigutía (BeAble Capital)
- María Ángeles Martín Prats (Skylife Engineering)
- José Ignacio Morales Conde (Telefónica)
Panel Session II - PhD graduates in industry: postdoctoral career development
The professional career of a PhD graduate in industry was reviewed: what is the role he/she has in the company, how does the company value his/her doctoral training, what are the most wanted skills. This time industry practitioners revealed interesting details of their postdoctoral experience.
- Erik Fernández Escudero (INEUSTAR)
- Adriana Gil Gil (CSIC)
- Rafael Romay Juárez (Teledyne-Anafocus)
- Manuel Domínguez Álvarez (ALTER Technology)
- Itziar Maestrojuán Biurrun (ANTERAL)
AI-DLDA 2020 International Summer School on Artificial Collaboration Session
Online from the University of Udine, June 29-July 3, 2020
AI-DLDA 2020 is the third edition of the International Summer School on Artificial Intelligence which is normally held in Udine, Italy. The team of the Department of Mathematics, Computer Science and Physics has organised this Summer School jointly with researchers from the University of Birmingham (UK) and the University of Modena and Reggio Emilia (IT). Following the development of the COVID-19 situation, this third edition of the Summer School has been held online.
This School has provided an in-depth analysis of the state-of-the-art research in the area of Artificial Intelligence, ranging from deep learning to data analytics. It has covered a large spectrum of current research activities and industrial applications dealing with big data, computer vision, human-computer interaction, biomedical informatics, deep learning, and serious games.
The School was organized in four teaching days with keynote lectures (during the morning) and laboratory sessions (during the afternoon). Topics covered by the lectures were:
- Cybersecurity and security
- Scene perception
- Food recognition
- Explainable AI
- Human behaviour analysis
The lecturers have been:
- Andrea Cavallaro (Queen Mary University of London)
- Andrew David Bagdanov (University of Florence)
- Davide Scaramuzza (University of Zurich and ETH Zurich)
- Emanuele Frontoni (Università Politecnica delle Marche)
- Luigi Di Stefano (University of Bologna)
- Petia Radeva (Universitat de Barcelona)
- Sebastiano Battiato (University of Catania)
- Wojciech Samek (Fraunhofer Heinrich Hertz Institute, Berlin)
ESRs of ACHIEVE-ITN participated in the School, organized by one of the PIs, Prof. Christian Micheloni.
Online Collaboration Session
Imasenic Zoom private, May 11, 2020
Because of the pandemic declared on March 11, 2020, all types of economic and social activities have been affected. Measures have been implemented for its containment at European, national and regional levels. In most of the countries of the participant institutions in ACHIEVE-ITN, periods of strict confinement and restricted mobility have been declared. All participants have been working remotely from home for large periods, therefore close interactions need to be promoted by the available means.
This online session is dedicated to explaining the status of the individual research projects (IRPs), their connection with the global plan. The main objective of the research programme is the development of a distributed vision platform composed of networked, smart and efficient embedded vision systems, that serves as the basic infrastructure for cooperative vision applications based on the in-node processing and extraction of relevant information. Interactions between IRPs are analysed and synergies are explored in order to conduct the work towards the common goal.
ESRs have presented their projects and have explored the connection points with their colleagues. Collaboration is initially centred in two clusters: one of groups dedicated mainly to hardware and the other of groups dedicated to software. Apart from this, lines for cooperation between hardware and software groups have been found during the session.
H2020 Marie Skłodowska-Curie Actions Artificial Intelligence Cluster Meeting
Brussels, December 10-11, 2019
The Research Executive Agency (REA) manages a large portfolio of H2020 projects and one of its roles is to provide programme and policy feedback on the funded actions to the European Commission. This time, a joint cluster event on artificial intelligence (AI) was organized together with the Commission's Directorate General for Education and Culture. MSCA units and researchers partnered with policy makers to provide participants a throughout scientific review and panel discussions. The objectives of the meeting were:
- showcasing the contribution of MSCA projects to AI R&I;
- promoting discussion and collecting information to provide coordinated input to the relevant EU policy-making services;
- enhancing synergies among projects and creating or reinforcing networking opportunities, particularly for MSCA fellows.
Project Coordinator, Ricardo Carmona, was invited to the panels concerning MSCA AI works in the DIGITAL world. He addressed the need for compact, lightweight and power-aware embedded vision systems, in order to develop distributed and autonomous artificial vision platforms for surveillance, monitoring of human activities and traffic management. To address privacy protection issues, this project follows a ‘privacy by design’ approach such that meta-data, rather than specific identification data, are being handled by the system.
International Conference on Distributed Smart Cameras
Trento (Italy), September 9-11, 2019
The 13th edition of the International Conference on Distributed Smart Cameras (ICDSC 2019) was held in Trento, Italy. It was organized by the Department of Information Engineering and Computer Science, University of Trento. This event was co-located with the 20th edition of ICIAP (International Conference on Image Analysis and Processing), organized biennially by the Italian Member Society (CVPL) of the International Association for Pattern Recognition (IAPR). Technical sessions were focused on/organized around:
- Cameras and Camera
- Face and Action Recognition
- Machine Learning
- Sensors and Architectures
- “Trends in Free Navigation Technologies” – Special Session
The last part of the program was conducted jointly with ICIAP. Audience from both events attended a shared technical session and a keynote by Prof. Davide Scaramuzza (University of Zurich) entitled “Autonomous, Agile, Vision-controlled Drones: from Frame to Event Vision”.
ICDSC’19 included sessions addressing innovative research in the broad field of camera networks and featured a PhD Forum and a Demo Session. ACHIEVE-ITN's students actively participated in the PhD Forum, presenting their latest results and discussing next steps in their research plan. Concerning the Demo Session, the students actively interacted with the showcased systems, which were directly related with the topics and research areas addressed in ACHIEVE. A total of 14 people involved in the Project participated in this activity.
Workshop on Architectures of Smart Cameras
Rennes (France), July 1-2, 2019
The 8th edition of the Workshop on Architectures of Smart Cameras (WASC 2019) was organised by Profs. Maxime Pelcat, Muriel Pressigout, and Karol Desnos at INSA-Rennes. This year’s topics covered the whole hierarchy of smart camera systems:
- smart image sensors
- heterogeneous image processing architectures
- embedded deep learning platforms
- software components
- embedded vision applications.
It was attended by 34 people. The occasion was the perfect setup for our students to develop their scientific communications skills. They presented their latest results and received some feedback from experts in theirs and complementary fields. Accompanying them, other young researchers from Univ. Santiago de Compostela, INSA-Rennes, Univ. Clermont-Auvergne and Univ. Seville presented their works.
A demo session was conducted by researchers from Edixia Automation, a company specialized in machine vision technologies based in the surroundings of Rennes. Edixia belongs to ID4CAR, a competitive cluster for the vehicle and mobility sectors.
Scientific course on WASC 2019 "Feature learning on embedded systems"
Rennes, (France), July 1-2, 2019
As part of the training program on complementary technical skills, a short-course was organised coincidentally with WASC 2019. The course was about feature learning in embedded systems, and was organized around this topics:
- Object detection and tracking based on deep learning
- Efficient implementation of CNN inference on FPGA
- NVIDIA platforms for embedded deep learning
The talks were centred in the embedding of deep neural networks both in adapted but closed architectures, like NVIDIA’s GPUs (Jetson Nano, TX1, TX2 and Xavier), and tailored architectures implemented on FPGA. It also covered examples of real-life applications in the traffic control and unmanned vehicle navigation fields. The instructors of the course were:
- Víctor Brea, from the Centre for Information Technologies of the University of Santiago de Compostela.
- Kamel Abdelouahab, PhD at the university of Clermont-Auvergne, and now with Sma-Rty, a Franco-Italian start-up on artificial intelligence and machine learning.
- Vincent Nguyen, Embedded Systems Architect from NVIDIA
A total of 34 people attended the course.
ACHIEVE-ITN Project Check Meeting
Seville (Spain), March 5, 2019
On M18 of the chronogram, the Project Check Meeting of ACHIEVE-ITN took place at the facilities of the Instituto de Microelectrónica de Sevilla (CSIC) in Seville, Spain. The event counted with the participation of representatives of all the beneficiary institutions. All the key people were present plus all the Fellows. The participation of the Project Officer was supported by videoconference because of logistic reasons. The main point of the agenda were:
- Address of the Project Officer to the Consortium
- Report of the Project Coordinator
- Introduction to the individual research projects of the ESRs
- Meeting of the Project Officer with the ESRs
- Meeting of the Project Officer with the IPs
Seminar on "Open Science"
Seville (Spain), March 6, 2019
On M18 of the chronogram, the Project Check Meeting of ACHIEVE-ITN took place at the facilities of the Instituto de Microelectrónica de Sevilla (CSIC) in Seville, Spain. The event counted with the participation of representatives of all the beneficiary institutions. All the key people were present plus all the Fellows. We seized the opportunity to celebrate two seminars on topics related with responsible research and innovation.
Open Science is a new approach to scientific research and dissemination based on cooperative work. It makes extensive use of digital diffusion technologies and collaborative tools. As part of our network-wide training programme, Ms. Isabel Bernal (General Manager of Digital CSIC) provided a thorough review of the concept.. She described the software infrastructure for reproducible and extensible research. The objective is to open the entire workflow, providing alternative evaluation methods and metrics, using open licenses, publishing open access, using open peer review, sharing, sharing preprints, using actionable formats, sharing protocols, notebooks, code, data, etc. After that, Isabel Bernal explained the H2020 mandate for Open Access publication and the basics of data management plans.
This seminar was attended by all ESRs and was open to students of the PhD program on Physical Sciences and Technologies of the University of Seville and the local scientific community in general. Total attendance was 30 people.
Seminar on "Ethics in Scientific Research and Publication"
Seville (Spain), March 6, 2019
Also belonging to the network training program on responsible research and innovation, a seminar on the ethical aspects around scientific research and publication was conducted. This seminar was divided in three parts:
The first one was led by Prof. M. Ángeles Oviedo Garcia, from the Dept. of Marketing and Market Research of the University Seville. The many facets of unethical and fraudulent scientific publication were treated. On the author side: plagiarism, data manipulation and data fabrication. On the editor and publisher side: coercive citation and predatory publication. At the end of the day, who protects the whistle-blower?
In the second part, we entered in the more general scope of ethics in scientific research with the help of Mr. Nico Formanek, from the High Performance Computing Center of the University of Stuttgart. In collaboration with Prof. Juan M. Durán —via Skype—, from the Dept. of Values, Technology & Innovation of TU Delft, Nico confronted the norms of science with the norms of industrial science. Ethics systematizes, provides foundation and recommends what is the appropriate response to intended and unintended consequences of our research. In the end, however, there is a personal and a societal view.
Finally, after the presentations, an open debate took place with participation of students and researchers. Total attendance was 30 people.
Demo session at ICDSC 2018
Eindhoven (Netherlands), September 3-4, 2018
2018 ACM International Conference on Distributed Smart Cameras was held at the High Tech Campus in Eindhoven. Members of the research group at IMSE (CSIC-Univ. Sevilla) teamed with scientists from Univ. Santiago de Compostela and the Politechnic Univ. of Cartagena to present results of their coordinated research on the integration of architectures and components for embedded vision.
Demos on 2D and 3D image sensors, high dynamic range, hardware platforms for embedded vision and deep learning and FPGA feature extractors were presented.
A total of 7 people involved in the Project participated in this activity.
Workshop on the Architecture of Smart Cameras
Coimbra (Portugal), June 27-28, 2018
The first technical activity organized within ACHIEVE-ITN was the 7th annual Workshop on the Architecture of Smart Cameras (WASC 2018). It was chaired by one of our partners, Prof. Helder Araujo from the Institute for Systems and Robotics (ISR) of the University of Coimbra. The workshop took place at the facilities of the ISR. It was attended by 40 people, including ACHIEVE-ITN participants and professors and students of the universities of Coimbra, Santiago de Compostela and Nantes. Technical sessions were organized around:
- Smart cameras and vision applications
- Smart image sensors
- Deep learning applications
- Smart cameras and image processing
WASC is dedicated to all aspects of embedded vision systems, smart camera networks, smart imagers and vision chips. The purpose of this workshop is to create a place to share experiences and concerns while presenting the latest advances and work in progress. The setup is rather informal in order to foster discussion and promote collaboration between researchers from complementary disciplines. The "motto of WASC" is defined by the absence of a registration fee, no parallel sessions, i. e. all attendees participate in the same activities, enough margin in the schedules to allocate discussions, and the promotion of demos, posters and full-interaction sessions.
Scientific course on "CMOS image sensors"
Coimbra (Portugal), June 27, 2018
A short course on CMOS image sensors was co-located with WASC 2018. It was attended by 40 people. This course is part of the complementary doctoral training on multidisciplinary topics around smart embedded vision systems. This short course on CMOS image sensors contained specific lectures on:
- Light detection in semiconductors
- CMOS image sensor architectures
- Noise in CMOS imagers
- Time-of-flight detection in CMOS
- Smart CMOS image sensors
The mission of this course was to provide a wide background on CMOS image sensors. Instructors for this course were members of the ACHIEVE-ITN consortium:
- Ángel Rodríguez-Vázquez (University of Seville)
- Ricardo Carmona-Galán (CSIC)
- Renato Turchetta (IMASENIC)
Seminar on "Sociological impact of surveillance technologies"
Coimbra (Portugal), June 27, 2018
Together with the development of core and complementary research skills, the ESRs are also trained in transferable skills that will help them to develop a highly employable professional profile. One of the aspects to be worked out at these network-wide training events is their role in carrying out responsible research and innovation. This seminar —also co-located with WASC 2018— is dedicated to the analysis of the effects of pervasive and continuous monitoring on society and human relations, from an ethical and from a legal point of view. Understanding these effects is the first step towards public engagement and the development of sustainable innovation in this field. This seminar covered aspects like:
- Privacy-aware monitoring
- Legal framework for surveillance
- Psychologic effects of constant monitoring
- Ethical use of surveillance technology
The seminar was delivered by Ms Marta Peirano, journalist and writer, deputy director of eldiario.es. It was attended by 40 people, including ACHIEVE-ITN participants and professors and students from the University of Coimbra.
ACHIEVE-ITN Kick-off Meeting
Brussels (Belgium), Nov. 8-9, 2017
This activity was realized during the kick-off at CSIC facilities in Brussels. Twelve people participated in the activity. All of the boards contributed to define the model of PCDP to be shared by the beneficiaries in order to coordinate and unify training criteria. This event did not involve the participation of ESRs, which had not been recruited yet. It was classified as a training event because of its influence on the complete programme.