vera.ai talk and asset presentation at the first ever AI-on-Demand webinar

vera.ai has contributed 12 assets to the AI-on-Demand platform. The platform's primary aim is to enhance accessibility to tools and solutions that support in combating disinformation. The assets developed by vera.ai partners and made available on the AI-on-Demand platform include datasets, ML models and a service. 

About the webinar

The webinar began with a brief overview of vera.ai's projects, highlighting its goals and objectives. It continued with a discussion of the challenges posed by disinformation and concluded with a showcase of the individual assets.

Individual assets

In detail, the asset contributions includes:

• three text-based ML models focused on general claim detection,
• two image-based ML models designed for detecting image forgery and synthetic images,
• three audio-related datasets, which include a dataset of synthetic speech, an audio phylogeny dataset and audio provenance dataset,
• a video-based service for keyframe selection and enhancement. 
• two datasets that support multimodal analysis, specifically addressing image-text decontextualization. 

The presentation featured an analysis of these assets, along with demonstrations of two specific examples: one for detecting image forgery and another one for identifying synthetic images, both integrated into the verification plugin (link plug-in to download page on Chrome store).

Here you can check out the assets and supporting information as they are included on the AI-on-Demand platform. 

Authors: Olga Papadopoulou (CERTH) with input from Martin Hyben (KInIT) and Luca Cuccovillo (IDMT)

Editor: Jochen Spangenberg (DW)