International AI Doctoral Academy (AIDA)

Author: Prof. Pitas Ioannis (The Aristotle University of Thessaloniki – AUTH)

The four ICT-48 networks (AI4Media, ELISE, HumanE-AI NET, TAILOR) and the VISION project joined forces and, under the joint initiative of AI4Media and VISION, founded a new joint instrument to support a world-level AI education and research programme.

The International AI Doctoral Academy (AIDA) has been created for offering access to knowledge and expertise and attracting Ph.D. talents in Europe. AIDA was very successfully launched on 3/11/2021. It has now 67 top AI Members (a very good mix of 50 excellent European Universities and 17 Research Institutes and Companies). AIDA is the first academy of its kind in Europe and internationally.

AIDA membership is excellent, its scope is pan-European/international and its aims are high to become a world reference for AI Ph.D. studies.

It is very important that 67 leading European AI partners joined efforts with the 5 Horizon Europe ICT48 AI flagship projects to foster Ph.D. education excellence in AI. This effort has strong momentum, as manifested by the 14 new Members that joined AIDA in the last 3 months. Therefore, AIDA can indeed attain a critical mass to have a large impact on AI academic education, industry workforce upskilling and in addressing very important social challenges, which range from the fight against disinformation to the provision of a human-centered and trustworthy AI that serves not only European citizens, but the humanity in general.

AIDA can also ensure European strategic autonomy in such critical technology like AI, with huge potential socio-economic impact and to reinforce Europe’s assets in AI, by benefiting its world-class researcher community, so that it stays at the forefront of AI developments. It can form a common AI resource center and become a shared facility offering access to knowledge and expertise and attracting talented researchers. It indeed aims high at becoming a world reference point, creating an easy entry point to AI excellence in Europe. AIDA is coordinated by the Aristotle University of Thessaloniki, Greece that is proudly named after Aristotle, the ancient Greek philosopher that founded Logic and Ethics, which both are at the core of AI.

Therefore, AIDA can boost a much-needed link between AI and humanities towards creating an anthropocentric (human-centered) European brand of AI that serves citizens worldwide. AIDA will also build a much-needed educational momentum to cater to the ever-growing societal and industrial needs towards building a strong, rich, human-centric and democratic Europe.

AI4Media’s integration with the European AI-on-demand platform

Author: Andreas Steenpaß (FRAUNHOFER IAIS)

The European AI-on-demand platform is a one-stop shop for anyone looking for AI knowledge, technology, tools, services and experts. The aim of this platform, which has been initiated by the AI4EU project, is to bring together the AI community while promoting European values and to facilitate technology transfer from research to industry. As a follow-up project of AI4EU, AI4Media is collaborating closely with the AI4EU platform by integrating the project’s outputs such as modules, services and algorithms into it as well as by organizing web cafés for community building. Due to these activities, AI4Media is one of the pillars for ensuring the sustainability of the AI-on-demand platform over the years to come.

In January 2019, the AI4EU consortium was established to build the first European artificial intelligence on-demand platform and ecosystem with the support of the European Commission under the H2020 program. As more and more features are integrated, the AI4EU platform serves as a catalyst to aid AI-based innovation, resulting in new products, services and solutions to benefit European industry, commerce and society. By bringing people together, the platform counterbalances the fragmentation of the European AI landscape.

Since the end of the year 2021 also marks the official end of the AI4EU project, it is now the task and the responsibility of the follow-up projects within the funding initiatives ICT-48 and ICT-49 to continuously animate the AI-on-demand platform by integrating new assets and features. The integration of AI4Media with the platform covers a wide spectrum of aspects which are reflected by the different sub-activities:

First, AI4Media ensures the publication of the AI resources developed within the project to the AI-on-demand platform. There is a large variety of types of resources such as service, dataset, docker container, library, or tutorial. All of them will be published online in the AI Catalogue. The high quality of the uploaded assets is guaranteed by the publication process, and each entry contains detailed information about the respective resource including a textual description, relevant documents, the license, and the GDPR requirements.

Second, AI4Media supports the community-building activities of the AI-on-demand platform by offering a series of live web cafés on AI. The goal of these sessions is to gain insights into the international AI scene, to share knowledge and experiences, and to meet stakeholders from various areas of AI research and application. The live web cafés regularly reach a very wide audience, while recordings of past web cafés are available on GoToStage. So far, there have been six sessions with contributions from AI4Media members and this exceptionally successful format will surely be continued.

Third, a selection of the resources published in the AI Catalogue is also technically integrated into AI4EU Experiments, an open-source platform for the development, training, sharing and deployment of AI models which constitutes the technical part of the AI-on-demand platform. Of course, this only applies to resources of those types where a technical integration is reasonable such as datasets and docker containers, but no tutorials. The selection is made based on the requirements of AI4Media’s use cases and on the impact of the relevant research.

Going beyond the publication and technical integration of AI resources, AI4Media will also provide showcases for the interoperability of AI4EU Experiments with other media platforms, which is a major success factor for wider dissemination on both sides. For example, it is foreseen to provide adapters for making modules from other platforms available in AI4EU Experiments.

Finally, the project also conducts in-depth research on the shifting approach from platform liability to platform responsibility for third-party infringing and/or illegal content. For this task, the focus lies on specific guidelines and recommendations regarding the impact of legal regulations on the AI-on-demand platform. It is worth emphasizing that some activities of the AI4EU project will continue under the umbrella of AI4Media, such as the very successful web cafés and the further development of AI4EU Experiments. Some of the aspects outlined above have also been discussed in much greater detail at the AI4Media workshop on the European AI-on-demand platform which took place on 11 November 2021 (Available on YouTube). Once again, the workshop has illustrated that the integration of AI4Media with this platform is a continuous process for the benefit of both projects.

Detecting deepfakes in multimedia content

Authors: Roberto Caldelli (MICC – University of Florence); Fabrizio Falchi (National Council Research – CNR); Adrian Popescu (French Alternative Energies and Atomic Energy Commission- CEA

When created by malevolent entities, deepfakes pollute the online space and have deleterious effects in users’ real lives, especially when aimed to interfere with debates related to polarizing situations. Deep learning has enabled the generation of credible deepfakes for different types of multimedia content, such as texts, videos and images. AI4Media proposes tools for efficient deepfake detection, regardless of the nature of forged documents.

Promising results were already obtained for texts and videos. Fake texts are difficult to distinguish from human-generated texts for short sequences. An efficient method was designed for the detection of fake tweets generated by specific accounts by learning adapted deep language models based per account. Deepfake videos are hard to detect when models are not specifically trained for a specific type of forgery. An algorithm that leverages the optical flow in videos was introduced and it successfully generalizes the detection capabilities to unlearnt forgeries.

Deep language models can be used to generate short texts, such as Tweets, which are difficult to distinguish from real tweets. Fake tweets are written by bots that mimic specific users by exploiting language models fine-tuned on the user’s past contributions.  The more refined the language models are, the more credible the generated fake tweets will be. The proposed fake tweet detection method is designed to match these generation practices and thus successfully distinguish fake from real tweets. A wide array of detection models were tested and the best results were obtained using a RoBERTa, a recent algorithm whose objective is to produce deep language models. It provides a detection accuracy of over 90%. The method can be used for an effective flagging of fake texts on Twitter. Importantly, it is easy to deploy for a large number of users who are of interest in AI4Media (or beyond) since language models are created per account. The work also led to the creation of TweepFake, a public dataset dedicated to the detection of deepfake tweets. The availability of this dataset will facilitate future research in the area and ensure the proposal of comparable and replicable results (Read the paper).

AI-technologies can be used in various ways to generate realistic fake videos. While the detection of known forgeries is well handled, the same is not true for forgeries that are not known to the detection algorithms and thus cannot be learned. The proposed model exploits the optical flow fields of the videos in order to improve the robustness of the detection of unlearnt forgeries. It also has competitive performance for the detection of learned forgeries. The main novelty is the integration of the bi-dimensional optical flow fields with the pre-trained network which usually receives the inputs of the three channels. This allows the detection of temporal inconsistencies which complement the information obtained from the usual frame-based analysis of content. This work paves the way toward the proposal of deepfake detection methods which are exploitable in practice since it generalizes to forgeries that are unknown and thus learned by detectors (Read the paper).

Work is currently ongoing to propose methods that combine different cues available in documents in order to seamlessly detect forgeries in multimedia documents. Early results obtained for deepfake videos which combine the visual and audio channels are particularly promising.


Caldelli, R., Galteri, L., Amerini, I., & Del Bimbo, A. (2021). Optical Flow based CNN for detection of unlearnt deepfake manipulations. Pattern Recognition Letters, 146, 31-37.

Fagni, T., Falchi, F., Gambini, M., Martella, A., & Tesconi, M. (2021). TweepFake: About detecting deepfake tweets. Plos one, 16(5), e0251415.

Building trust in Artificial Intelligence – AI4Media’s contribution to an ethical AI

Author: Killian Levacher, Research Scientist (IBM Research Europe – Dublin)

One of the research domains where AI4Media has focused on is the critical infrastructure necessary for the inclusion of AI tools within our society by investigating the various dimensions of Trusted AI. During the first year of the project, our activities have already successfully provided various research contributions in areas such as AI Robustness, AI Fairness, Explainable AI and AI Privacy.

This work has also led to the publication of 6 papers in prestigious AI conferences, and the submission of 4 conference papers and 1 journal paper. These early accomplishments represent a solid foundation to expand our research throughout the remaining years of the project.

Artificial Intelligence (AI) is an area of strategic importance to the European Union with respect to its ability to support and shape future economic and social development. While the recent leaps in innovation in this space offer immense opportunities, due to the increasing importance and prevalence of AI systems across industries various aspects of this technology present many security, as well as, societal risks which may conflict with the ethical and democratic principles shared across the European Union such as transparency, privacy and inclusion among others.

Trustworthy AI hence aims at providing a framework for the development of Machine Learning (ML) technologies, which guarantees their suitability with respect to the democratic and ethical values shared in our society. This recently emerging field of AI can be typically divided into four broad dimensions, namely AI robustness, Explainable AI, AI fairness and AI privacy.

AI Robustness focuses on machine learning vulnerabilities that can be exploited by malicious attackers seeking to either steal capacities of proprietary models, identify private information used to train these models, or purposely push a model in making incorrect predictions. These attacks can be achieved through the use of adversarial samples in various forms (images, texts, tabular data, etc.) and across a wide range of model types. In the first year of the AI4Media project, our activities already successfully provided various research contributions in this field.

  • The Aristotle University of Thessaloniki was successful in creating a novel AI training method which uses hyperspherical class prototypes to increase neural network robustness of AI models. (Read the paper)
  • IBM discovered that deep generative models can also be vulnerable to a new set of backdoor attacks. As part of this work, IBM developed new defence capabilities to protect generative AI models against such attacks. (Read the paper)
  • A new attack algorithm promoting the robustness of re-identification systems was developed by the University of Trento which addresses known vulnerabilities of such systems when used in domains unseen during their training phase. (Read the paper).

Explainable AI deals with the trust that needs to be established between an AI model and its user. European legislation states that technical measures must be put in place in order to facilitate the interpretation of the outputs of AI systems by the public. In other words, users of AI models must be able to understand why predictions were made, regardless of the precision or validity of each prediction. While the recent explosion of deep learning models has led to amazing gains in performance, these models in particular provide very limited visibility even to their own designers as to how they reached a decision. It is, therefore, crucial to develop a set of technologies that can support users in understanding how specific predictions were made, in order for these technologies to be safely incorporated within the fabric of society.

During the first year of the project, AI4Media partners successfully made a few contributions in this dimension of Trusted AI.

  • The University of Applied Sciences and Arts Western Switzerland developed a new method which enables the public to understand and interpret the most salient internal features of deep learning models in order to understand why a specific decision was made. The Centre for Research and Technology Hellas on the other hand developed a technique which can identify within videos, the most important items present within a frame which should be used by AI models to explain and describe a specific scene to the public (Read the paper).
  • The Universite Cote D’Azur developed a novel method for providing explanations for decision-marking systems.
  • The Commissariat `a l’Energie Atomique et aux Energies Alternatives proposed a new technique which enables vector arithmetic to be used in the underlying process used by generational models to produce new synthetic material.

Thanks to the collaboration of various partners, a public workshop (available on YouTube),  dedicated to developing a taxonomy of Explainable AI across various disciplines was also organised, bringing together 16 experts (7 invited speakers and 6 invited panelists) from a wide range of disciplines (technologists, philosophers, lawyers etc.) to discuss the various meanings, legal constraints and social impacts of Explainable AI and how these will impact the future technical development of the field.

Finally, the process of training and building AI models requires the management of large amounts of data which in many cases contain sensitive information which should not be shared beyond a dedicated group of data processors and owners. This generates a conflict of interest between the need to have the most numerous and accurate data available to reach high precision accuracy while at the same time reducing the amount of data being used to minimise any impact on an individual’s privacy.

Private information leakage can occur both while a model is being trained as well as after deployment. AI Privacy hence aims at threading the needle between these two forces by providing the means to produce reliable ML models while simultaneously protecting individuals’, as well as, corporations’ sensitive information. In this domain, during the first year of the project, the IDIAP Research Institute developed a new tool to secure privacy within a specific type of neural network based on graphs. (Read the paper)

This resulted in a publication on “Locally Private Graph Neural Networks” which was shortlisted as one of the ten finalists for the CSAW 2021 Europe Applied Research Competition. This competition awards the best paper of the year written by doctoral students in the field of security and privacy. A differential privacy library for AI models (Access the GitRepo), was developed by IBM and a novel method for data protection using adversarial attack was developed by the Aristotle University of Thessaloniki (Read the paper).

The use of AI in the media sector: policy and legislative developments at the EU level

Author: Lidia Dutkiewicz, Legal researcher (Center for IT & IP Law (CiTiP), KU Leuven)

During the past year, the European Commission (EC) proposed a comprehensive package of regulatory measures that address problems posed by the development and use of AI and digital platforms. These include the AI Package, the Digital Services Act and the Digital Markets Act, as well as the Data Governance Act and the forthcoming Data Act. In particular, the Artificial Intelligence Act (AI Act) presented by the EC in April 2021 represents a key milestone in defining the European approach to AI.

In a first of its kind legislative proposal, the EC aims to set a global standard of how to address the risks generated by specific uses of AI through a set of proportionate and flexible legal rules. The key question is how will these legislative proposals affect the use of AI in the media sector?

During the first year of the AI4Media project, one of the key milestones was to acheter sildenafil 100mg provide a clear overview of existing and upcoming EU policy and regulatory frameworks in the field of AI. In the last few years, there has been a variety of publications, guidelines, and political declarations from various EU institutions on AI. These documents provide valuable insight into the future of AI regulation in the EU. However, the large number of developments in the EU in the area of the “AI policy initiatives” makes it very difficult for AI providers and researchers to monitor the ongoing debates and understand the legal requirements applicable to them. The key challenge is to assess the possible implications of the proposed rules on AI applications in the media sector, i.e. for content moderation, advertising, and recommender systems.

Our analysis of the EU policy on AI envisages the impact of these EU initiatives for the AI4Media project in four distinctive areas. First, access to social media platforms’ data allows researchers to carry our public interest research into platforms’ takedown decisions, recommender systems, mis- and disinformation campaigns and so on. However, in recent years it has become increasingly difficult for researchers to access that data. That’s why there is a clear need for a legally binding data access framework that provides independent researchers with access to a range of different types of platform data. Recent regulatory initiatives, such as the Digital Services Act (DSA) try to address this problem.

Article 31 of the DSA proposal provides a specific provision on data access. However, it narrows access to platforms’ data to “vetted researchers”, namely university academics, which excludes a variety of different actors: journalists, educators, web developers, fact-checkers, digital forensics experts, and open-source investigators. Moreover, “vetted researchers” will be able to access platforms’ data only for purposes of research into “systemic risks”… The final scope of this provision will, undoubtedly, shape the way in which (vetted) researchers, journalists, and social science activists will be able to access platforms’ data. This is particularly relevant for the AI4Media activities such as opinion mining from social media platforms or detection of disinformation trends.

Second, it is extremely important to clarify the position of academic research within the AI Act. It is currently unclear whether the AI Act’s primary objective i.e. “to set harmonised rules for the development, placement on the market and use of AI systems” and its legal basis exclude non-commercial academic research from the scope of the Regulation or not.

Third, the scope of the AI Act is unclear when it comes to its applicability to media applications. Importantly, certain practices such as the use of subliminal techniques or the use of the AI system which exploits the vulnerabilities of a specific group of persons, are prohibited. However, the current wording of these provisions makes it unclear whether and to which extent the online social media practices such as dark patterns fall within the scope of this prohibition. The AI Act also proposes certain transparency obligations applicable to AI systems intended to interact with natural persons, emotion recognition systems and deep fakes. However, the requirements lack precision on what should be communicated (the type of information), when (at which stage this should be revealed) and how. The important research questions which will be tackled in future AI4Media activities include: 

  • Does the ‘AI systems intended to interact with natural persons’ encompass recommender systems or robot journalism?
  • Does ‘sentiment analysis’ and measuring and predicting user’s affective response to multimedia content distributed on social media with the use of physiological signals fall under the “emotion recognition” system?

Fourth, the use of AI to detect IP infringements and/or illegal content is one of the key legal and societal challenges in the field of AI and media. The key questions center around the role of ex-ante human review mechanisms before removing content and the potential violation of human rights, i.e. freedom of expression when legal content is being removed. The platform responsibility for third-party infringing and/or illegal content will be particularly relevant in AI4Media’s activity related to the “Integration with AI-On-Demand Platform”.

Considering the ever-changing legal landscape, the work performed in the analysis of the EU policy and regulatory frameworks in the field of AI is not a one-off exercise. Rather, the preliminary analysis done so far serves as a solid basis for the upcoming work in the later stage of the project, namely “Pilot Policy Recommendations for the use of AI in the Media Sector” and “Assessment of social/economic/political impact from future advances in media AI technology and applications”.

Stay tuned for more!

AI4Media Workshop on the European AI-on-demand platform

On November 11th, 2021, AI4Media organised a workshop on the European AI-on-demand platform.

The objective of this workshop was to allow a better understanding of the technical and non-technical facets of the AI-on-demand platform, highlighting the role of the platform as the central link between the European AI networks, and to encourage everyone interested in AI to join it.

The AI-on-demand platform has been initiated by the AI4EU project, and it aims to bring together the AI community while promoting European values, and facilitating technology transfer from research to business.

As a follow-up project of AI4EU, AI4Media is collaborating closely with the AI4EU platform by integrating the project’s outputs such as modules, services and algorithms into it as well as by organizing AI4EU Web Cafés for community building.

The workshop was divided into two parts, the first one was dedicated to presenting the platform, including the organisations and projects involved and its different parts such as the AI Catalogue and the web cafés. During this session it was also illustrated the cooperation between European AI networks on the AI-on-demand platform, with special emphasis on digital innovation hubs.

The second part was focused on presenting the AI4EU Experiments, which is an open-source platform for the development, training, sharing and deployment of AI models which constitutes the technical part of the AI-on-demand platform. This included the introduction to general features such as the Marketplace and the Design Studio as well as some example pipelines. How AI4EU Experiments can be connected to other media platforms and how new modules can be integrated into it, was also shown in a tutorial in this session.

Workshop Recording





AI4Media may use cookies to store your login data, collect statistics to optimize the website's functionality and to perform marketing actions based on your interests.

They allow you to browse the website and use its applications as well as to access secure areas of the website. Without these cookies, the services you have requested cannot be provided.
These cookies are necessary to allow the main functionality of the website and they are activated automatically when you enter this website. They store user preferences for site usage so that you do not need to reconfigure the site each time you visit it.
These cookies direct advertising according to the interests of each user so as to direct advertising campaigns, taking into account the tastes of users, and they also limit the number of times you see the ad, helping to measure the effectiveness of advertising and the success of the website organisation.

Required Cookies They allow you to browse the website and use its applications as well as to access secure areas of the website. Without these cookies, the services you have requested cannot be provided.

Functional Cookies These cookies are necessary to allow the main functionality of the website and they are activated automatically when you enter this website. They store user preferences for site usage so that you do not need to reconfigure the site each time you visit it.

Advertising Cookies These cookies direct advertising according to the interests of each user so as to direct advertising campaigns, taking into account the tastes of users, and they also limit the number of times you see the ad, helping to measure the effectiveness of advertising and the success of the website organisation.