The EU co-funded InVID project (2016-2019) focussed on the algorithm-supported analysis and verification of videos, something its predecessor REVEAL wasn’t able to cover. Not only did we learn a lot about manipulated video content and ways to detect it in the course of the InVID effort, the consortium also managed to create a number of useful tools for the international verification community. Here’s an overview.
The InVID Verification plug-in (browser extension)
The InVID plug-in supports the analysis and verification of images and videos. A primary aim of the service is to debunk misleading or manipulated visual content. The idea behind the plug-in was to bring together a collection of different tools, services and open source libraries to provide users (primarily journalists and human rights abuse investigators) with a comprehensive toolbox in a single user interface. The development was led by AFP (Agence France Presse),
The InVID plug-in provides services that allow users to
- extract contextual information from YouTube videos, Twitter videos and Facebook public videos (analysis feature)
- fragment almost any video into keyframes in order to perform an image reverse search on the resulting individual images (keyframes feature);
- query thumbnails that are provided automatically (and randomly) by YouTube, through several reverse image search engines (thumbnails feature)
- an advanced Twitter search that allows for searches exact to the minute
(Twitter search feature)
- enlarge parts of images with a magnifying lens in order to “zoom in” on details of an image and possibly discover further information contained in these images or video keyframes (magnifier feature)
- extract and read metadata for both still images and videos (metadata feature)
- negotiate usage rights for a video and inform users about what they can do with it should they wish to re-use and/or publish content that they come across on YouTube, Facebook and Twitter (video rights feature)
- perform a forensic analysis of still images (forensic feature; note: this was developed by DW and CERTH-ITI within the scope of REVEAL)
The InVID plug-in is in use in many newsrooms around the world, but also at human rights organizations. At the time of writing this text, (October 2019) there are more than 3,000 regular users of the browser extension. The consortium feels that it has provided an important service – something that is “here to stay”.
The InVID Verification App
The InVID Verification App is a desktop-based service that targets the professional video analysis market, such as newsrooms. It supports the verification of videos by providing a user interface that takes into account the verification process – from content analysis via tracking of verification results to the final assessment stage. The interface provides features for playout, an analysis of individual aspects contained in the entire video, and allows for a deeper examination of relevant video shots and subshots all the way to individual frame level.
The InVID Verification App is provided by Condat AG. It was developed under the guidance of InVID project partners AFP and DW. Presently, a free version is provided for the Chrome, Firefox and Opera Browser. Under http://invid.condat.de/ you can browse through videos that have already been analysed, so they can get an idea of the tool’s features. You can also analyse your own/new videos by logging into the system with a Google, Twitter or Facebook account, or create a new and independent identity.
Similar to the verification plug-in, the InVID Verification Application also integrates a set of components related to the analysis and rights management of user-generated digital content. These, in turn, were developed by project partners CERTH, Universitat de Leida, eXoMakina and MODUL Technology.
The free online version was launched in March 2019 and had gained more than 8000 unique users by September 2019.
The Visual Analytics Dashboard
The Visual Analytics Dashboard, developed by InVID project partner webLyzard, is a search and information exploration platform that allows for the tracking and visualization of evolving stories across multiple web sites and social media platforms. The system identifies opinion leaders who trigger and amplify a story, and provides visual tools to analyze the relations among these opinion leaders.
The dashboard uses the multiple coordinated view technology for the desktop version, and a cross-platform HTML5 application to access analytic functions through smartphones and other mobile devices. The dashboard’s real-time synchronization mechanisms allow for the tracking of information flows within InVID’s contextualized information space. This includes the ability to display image and video content (using thumbnails to represent related stories and content clusters), thereby integrating visual content into existing web intelligence and knowledge co-creation workflows.
The final version of the dashboard, as it was developed within the InVID project, supports:
- story detection across social media channels
- story-based newsworthy video identification
- automatic metadata extraction and indexing
- content exploration and visualization
- one-click submission of selected videos to the InVID Verification Application
- rendering of the geographic distribution of emerging stories
- embedded playback at the video and fragment level
- automated PDF reporting
If you’re interested in testing the dashboard, please get in touch with webLyzard.
Verification and tools beyond InVID
InVID was recently completed and wrapped up, but its results will of course be put to good use in other projects, e.g. WeVerify, which is also funded by the EU. WeVerify will ensure that the InVID browser extension lives on. We even get the chance to further enhance it and continue with its development. There is, in fact, already a new version of it, the co-branded InVID WeVerify verification plug-in.
Thanks to WeVerify we also get the opportunity to expand our collaborative verification platform Truly Media; individual WeVerify features and services are likely to be integrated there. Stay tuned!
We’re very happy with the progress we’re making in the field of verification. InVID let us take a big step forward here.
More links and resources
Some consortium members wrote a book about video verification and related InVID work. It’s called Video Verification in the Fake News Era and IS available via Springer Publishing.
Author: Jochen Spangenberg (using information and screenshots provided by the InVID Consortium). Key visual by freestocks.org on pexels.com