AIMultimediaLab

Multimedia Evaluation Benchmark MediaEval2025 https://lnkd.in/dam7CQrC -- The MediaEval Multimedia Evaluation benchmark offers innovative challenges that are of interests to researchers in the areas of , , , , . Registration is now open and submissions will be in October.

This year’s tasks are:
- : VQA (with multimodal explanations) for gastrointestinal imaging
- : Predicting movie and commercial memorability
- : Multimodal summarization of multiple topically related websites
- : Retrieval and generative AI for news thumbnails
- : Advancing detection of generative AI used in real-world online images

MediaEval places special emphasis on gaining insight into data and algorithms and moving forward the state of the art on tasks that are novel or unexpectedly challenging, due to the nature of the data or to a limited quantity of labels. The overall mission of MediaEval is to support that makes multimedia a positive force for society.

Results are presented at the next edition of the yearly MediaEval workshop, which will take place in Dublin, Ireland, Sat.-Sun. 25-26 October 2025, between CBMI 2025 (the 22th Content-Based Multimedia Indexing) and ACM Multimedia 2025 (the 33rd ACM International Conference on Multimedia). The workshop will provide an opportunity for online participation.

4th ACM International Workshop on against , organized with ACM , Chicago, USA, June 30-July 3, 2025 https://mad2025.aimultimedialab.ro/ April 10, 2025.

Topics of interest (not limited to): Disinformation detection in multimedia content (e.g., video, audio, texts, images), Multimodal verification methods, Synthetic and manipulated media detection, Multimedia forensics, Disinformation spread and effects in social media, Analysis of disinformation campaigns in societally-sensitive domains, Robustness of media verification against adversarial attacks and real-world complexities, Fairness and non-discrimination of disinformation detection in multimedia content, Explaining disinformation detection results to non-expert users, Temporal and cultural aspects of disinformation, Dataset sharing and governance in AI for disinformation, Datasets for disinformation detection and multimedia verification, Open resources, e.g., datasets, software tools, Large Language Models for analyzing and mitigating disinformation campaigns, Large Multimodal Models for media verification, Multimedia verification systems and applications, System fusion, ensembling and late fusion techniques, Benchmarking and evaluation frameworks.

See the previous edition proceedings here https://dl.acm.org/doi/10.1145/3652583.3660000.

Project support UEFISCDI DeteRel SOL12/2024, AI4Debunk, vera.ai, news-polygraph.

Universitatea POLITEHNICA din București, Università Mercatorum, Fraunhofer Karriere IDMT, České vysoké učení technické v Praze, Εθνικό Κέντρο Κοινωνικών Ερευνών - ΕΚΚΕ CERTH, CEA-List, Technische Universität Berlin.

23th edition Multimedia Retrieval in CLEF https://lnkd.in/dYVWswdU:
📀 (9th edition) concept detection & caption prediction in radiology medical images https://lnkd.in/dxtKkhmd
🎞️ (3rd edition) traceability of training data in synthetic lung tuberculosis image generation https://lnkd.in/dmb32Ncc
📼 (2nd edition) multimodal and generative TelemedICine https://lnkd.in/dTAmUecj
💾 (3rd edition) visual question answering & synthetic data for enhance diagnostic in gastrointestinal images https://lnkd.in/djUtkZW3
💽 (2nd edition) translating text or speech into sequences of pictogram terms linked to ARASAAC https://lnkd.in/dCSxNtNm
💿 (new) LLMs' reasoning capabilities on complex inputs on student exam data https://lnkd.in/dVktcZQ2
📖 (joint task) develop means for automated finding suitable images for arguments https://lnkd.in/ddvucU49
🚀 ***Run submission is due May 10, 2025*** https://lnkd.in/dB97tVDf ImageCLEF AI4Media Universitatea POLITEHNICA din București AIMultimediaLab HES-SO

Check out our latest results in generating medical images via a latent-to-image approach in conditional . Apart from the high quality of the results, i.e., doctors cannot differentiate the generated vs real images, our approach ensures -preserving capabilities, i.e., synthetic medical images tend to contain traces of the original real training images, which may lead to revealing the identity of the source – see A.-G. Andrei, M.G. Constantin, M. Graziani, H. Müller, B. Ionescu, “Privacy Preserving Histopathological Image Augmentation with Conditional Generative Adversarial Networks”, Pattern Recognition Letters, Impact Factor 3.9, 2025 https://lnkd.in/dzurYV69 Universitatea POLITEHNICA din București HES-SO Valais-Wallis.

Congratulations to Alexandru Ioniță for winning the 2x 1st place (both in individual and team participation 😃) at the IBM Quantum: Qiskit Fall Fest https://lnkd.in/gdakvYuZ which was organized with the Military Technical Academy "Ferdinand I" -- Qiskit Fall Fest is a collection of events that invites students, researchers and industry professionals around the world to participate in a wide array of quantum-themed activities, ranging from quantum challenges, hackathons, and coding competitions to workshops, social events, and more. PS: Alexandru is developing his PhD thesis on quantum Artificial Intelligence. Keep up the great work! Universitatea POLITEHNICA din București

our special issue on Realistic Data: , , organized with ACM Transactions on Multimedia Computing, Communications and Applications, guest editors Bogdan Ionescu, Ioannis Patras, Henning Müller, Alberto Del Bimbo, is now online https://lnkd.in/dDtWR9Es – it is a timely collection of contributions targeting the various facets and modalities, e.g., signals, images, volumes, audio, etc., of the current trends in generating synthetic data, i.e., data which looks as real as it gets but it is computer generated. We received a record number of 40 submissions out of which 15 were finally accepted for publication. Enjoy your reading! Universitatea POLITEHNICA din București, Queen Mary University of London, HES-SO, Università degli Studi di Firenze.

… and it started, the 1st on for https://www.aimultimedialab.ro/1st-workshop-on-artificial-intelligence-for-multimedia/ -- the yearly workshop of the AI Multimedia Lab https://www.aimultimedialab.ro/. The workshop is dedicated to the PhD students of our lab who are presenting in front of the other colleagues, with the objective of sharing ideas and results as well as brainstorming. Wonderful people doing science! Campus Research Institute Universitatea POLITEHNICA din București

We were there Congress of the University of Medicine and Pharmacy "Carol Davila" Bucharest – congratulations to Lucian Gruia for presenting our work on and for Enhanced Research and Discovery – is key to fostering breakthroughs in technology for the good use of society – warm thanks for the invitation and looking forward to the next year’s edition! Doctoral School of Electronics, Telecommunications & Information Technology Universitatea POLITEHNICA din București

A new doctor is born – congratulations to Ciprian Zamfirescu for successfully defending his PhD thesis awarded with the highest distinction “Summa cm laude” in engineering – this is a one of a kind contribution as you balanced the & with via deploying your knowledge to undergraduate students, via deploying your knowledge to products in Vodafone Romania, and via deploying your solutions to public use, e.g., to Salvamont Romania-Dispeceratul National Salvamont for the UAV fleet which saves lives, and popularizing the technology through countless public presentations, e.g., https://lnkd.in/dnqPkGWu – it resulted the https://lnkd.in/dDTikusj with where hundreds of students and of your (and teams') results each year ... and this is just the beginning of the journey! Many thanks to the PhD defense committee Prof. Corina Naforniță, Prof. Emanuel Pușchiță, Prof. Ion Marghescu, and Prof. Mihai Ciuc Doctoral School of Electronics, Telecommunications & Information Technology Facultatea de Electronică, Telecomunicații și Tehnologia Informației Universitatea POLITEHNICA din București.

Check out our latest release on generating of flow-based – this is a COVID pandemic triggered application, i.e., helping to optimize and test network traffic via data that replicate very closely the actual behavior of home/office users, e.g., using software tools, social media, file transfer and so on. We also tested the data generator in augmenting data for traditional machine learning applications and successfully boosted . The data and code are – see Mihai Gabriel Constantin et al., “Exploring Generative Adversarial Networks for Augmenting Network Intrusion Detection Tasks”, ACM Transactions on Multimedia Computing, Communications and Applications, Impact Factor 5.2, Q1, https://dl.acm.org/doi/10.1145/3689636 – this is the outcome of a wonderful cooperation and industry project with Keysight Technologies Romania & USA Universitatea POLITEHNICA din București.

We were there (25th edition) https://clef2024.imag.fr/ Grenoble, France -- great job Alexandra Andrei! -- presenting our annual ImageCLEF lab https://lnkd.in/dVWu595z -- in particular, we were leading the overall lab and the task which this year fostered two challenges: (i) Identifying training data “fingerprints”, and (ii) Detecting generative models’ “fingerprints” -- in numbers, 35 groups submitted >220 runs, 35 working notes were published -- another successful year having another stepping stone laid out Universitatea POLITEHNICA din București HES-SO.