Results

Services & Tools 

Here you will find descriptions, demo videos and, created code created and published as a result of the MediaVerse project.

The current version of the MediaVerse Node is available for download under the Apache 2.0 Licence on github. You will find a detailed description here.

Download the MediaVerse Node from GitHub

Asset Management: MAAM

MediaVerse Asset Management and Moderation (MAAM): a fork of the MV node and its core architecture, with an extensive redesign and re-implementation of the back-end and front-end for more optimized user experience and better performance. The focus is on managing the automatic and user-provided annotations to the media assets, and serve as a basis for further integrations such as forensics analysis. 

For more information visit https://maam-next.mever.gr 

MAAM is available as a service. Please contact Symeon Papadopoulos at CERTH

Accessibility Toolset

MediaVerse’s accessibility toolset includes advanced features to comply with regulations and guidelines for people with disabilities. Key functionalities include speech-to-text with multilingual translation and a post-editor. These features enhance the quality and accuracy of subtitles and allows to manually generate captions, making content more accessible to a diverse audience.

Find out more about this product at https://webs.uab.cat/livesubtitling/nerstar/and  in the MediaVerse YouTube channel

Accessibility: RACU - Subtitling Workflow

RACU stands for “reconnaissance automatique avec corrections ultérieures” (automatic recognition with retrospective corrections). Partner STXT provides the RACU workflow as an accessibility tool for the MediaVerse project. It complies with the latest regulations and guidelines for people with disabilities and can be used to transform different types of content to be accessible.

The process starts with automatic speech recognition of the uploaded videos. Afterward, the text is machine translated and automatically segmented into subtitles. The multilingual subtitles are made available as SRT or VTT files for download.

Asset Management: Content Recommendation

Partner LINKS Foundation developed a content-based recommendation system to facilitate content discovery and retrieval within the MediaVerse network.

The content-based recommender system allows content discovery by exploiting the latest uploads of the user and the retrieval of assets similar to the ones that the user selects.

This Open Source tool is available under the Apache 2.0 licence.

Content Moderation Toolset

MediaVerse includes a  content moderation toolset, which comprises detection models for: a) misleading, or false media content, b) content associated with hate speech and cyberbullying, c) disturbing content including violent and gruesome scenes, d) nudity, profanity, and other types of NSFW content.

For more information on this proprietary tool, please contact Symeon Papadopoulos at CERTH.

Content Moderation: Near-Duplicate Detection

The Near-Duplicate Detection (NDD) service is an online tool created by CERTH for the detection of highly similar or near-duplicate multimedia assets, i.e., images, videos, and audio, from large-scale databases given a query. NDD facilitates the identification of the originality of the new assets submitted to the MediaVerse platform, contributing to the copyright management of the multimedia assets. To this end, our tool provides carefully crafted functionalities for indexing and search of multimedia assets based on state-of-the-art algorithms produced from our extensive research on the field of Near-Duplicate Detection. It constitutes a fast and accurate solution designed to support the distribution of processing to the network of MV nodes and facilitate federated search.

Watch this short demo video to see our tool in action.

Find out more about this open source tool at https://mever.iti.gr/video_search/ and https://github.com/mever-team/distill-and-select 

Immersive Storytelling: FADER 360° Experiences

Fader is an authoring tool for users without programming expertise to produce high-quality interactive 360° experiences.

Partner Vragments has improved 360º production processes in newsrooms by assessing and further optimizing technical elements of the process in the backend, including:

  • Swift transcoding of 360° video files,
  • 360° repository for collaborative work.

You can try out some of the features here and find first MediaVerse demonstrators in our blog (with links to interactive demos):

Immersive Storytelling: Scrollmersive - 3D VR Scrollytelling

Our partner Vragments examines in how far spatial storytelling can be integrated into an automatic production process at newsrooms. Allowing users to navigate through a narrative by scrolling from one 3D story point to the next adds value to any storyline. More information can be found here.

Immersive Storytelling: VRodos 3D plugin for WordPress

Our partner CERTH investigates the possibility of automating the visualization of 3D models in WordPress by allowing the upload, configuration, and visualization of 3D models on the front-end of WordPress sites without coding. 

The VRodos plugin allows artists to make their own a repository within WordPress for managing their assets. In this manner, they become independent of central repositories and they can visualize their content in their selected context.

You can find the VRodos plugin in Github.

Immersive Storytelling: VRodos Authoring Tool

VRodos ia a novel, collaborative, authoring space which incorporates tools for avatar creation, creation of VR personal spaces while simultaneously enabling social interactions.

The VRodos Scene Editor allows to combine assets in 3D scenes. It is characterized by four main interaction areas, the preview area that covers the entire background, the scene and object controls, the asset inventory, and the scene manager.

The preview area is a real-time authoring environment to facilitate basic actions such as transformation, rotation, and scaling of 3D models with widgets. The inventory of assets allows the author to browse assets and drag-n-drop them into the scene. It also offers the link to the Assets Editor for adding new assets or editing an existing one. Multiple instances of an asset can be inserted into the same scene.

In the preview area other actions can be performed, such as object re-positioning, re-scaling, and rotation angle determination. Right-clicking on an object allows to change certain properties, e.g., doors have as properties the name of the scene where the user will be transferred upon exiting that door.

The scene manager allows to create/edit/delete a scene as well as navigate across them.

Media Annotation: Automatic Tagging

The MediaVerse Node’s Asset Management functionality includes tools for the extraction of tags, metadata, and semantic representations of new media types (3D, 360, images, TikTok videos, gifs, memes).

This Open Source tool is available under the Apache 2.0 licence.

Media Annotation: TransNER Text Annotation Service

TransNER is an annotation tool for text supporting the extraction of various types of entities based on AI methods.

This Multilingual Tool (English, Italian, Dutch) finetuned on News Article comes from Wikiner. It annotates texts with one of the following entities: PERSON, LOCATION, ORGANIZATION, MISCELLANEOUS. It allows to better support text semantics identification.

You can find this Open Source tool (Apache v2.0) at https://github.com/D2KLab/transner

Media Annotation: Media Annotation Service

MediaVerse include a proprietary service for annotating assets with tags and metadata, and extracting  semantic representations of new media types.

The media annotation service offers automatic captioning, object detection, action recognition, face detection/recognition and moderation models, such as classifiers for NSFW and disturbing content. In addition, a vector representation that encodes the semantic information of visual content is provided to help with retrieval.

FOr more information, contact the coordinator at CERTH: Symeon Papadopoulos, papadop@iti.gr

Social analytics engine

The main functionality of the Social Analytics Engine are essentially: content ingestion from owned social media accounts for the enriching of a MediaVerse node content; Content publishing in social media platforms; and performance tracking of shared content, including trending topic detection.


For more information on this prorietary tool, please contact Tasos Lampropoulos at ATC.