Artificial intelligence (AI) adoption has certainly been through the technology hype cycle, with expectations often higher than today’s algorithms can meet. But AI usage is growing steadily, particularly in specific areas such as data management and automation.
Indeed, a recent survey from IDC found that while AI/ML initiatives are steadily gaining traction (31% of respondents have AI in production), most enterprises are still in an ‘experimentation’, ‘evaluation/test’, or ‘prototyping’ phase. However, respondents still widely believe that the top three benefits of AI are: improving customer satisfaction, automating decision making, and automating repetitive tasks.
These last two benefits are particularly relevant to enterprise Digital Asset Management (DAM) platform adoption. Moreover, they are helping more enterprise customers to enrich their media collection than ever before. So much so, ‘knowledge management’ is predicted to be the fastest expanding area of AI in 2022, according to Gartner. Its analysts believe that, while worldwide AI software revenue will see a robust year-on-year increase of 21.3% in 2022, the knowledge management segment will see a massive 31.5% uplift.
A key challenge for enterprise-grade DAM is applying high-quality metadata for images at scale. While manual processes can result in very high-quality metadata, older (and non-AI-based) automated processes can result in inaccurate tag generation, making assets more difficult to find and retrieve from archives.
Accurate automated metadata can add additional value, far beyond the practical abilities of manual processes. From automatic transcription tools to sponsor logo detection, AI can enrich your media library in multiple ways. These not only help you save time and reduce human error, but to make the most of your existing data and get projects to market faster.
1. Automatic speech-to-text output
Manual transcription is fortunately a thing of the past. But even the best modern text-to-speech algorithms aren’t perfect, forcing some level of human oversight, editing and sign-off. That said, by augmenting speech to text tools with AI, the matching success rate can be enormously improved, reducing the human input required.
The accuracy of AI today can also be invaluable in situations where distortion or background noise makes more traditional transcription tools so error prone as to be almost useless. This can help bring to life legacy footage that was simply too onerous to transcribe manually, as well as create new content associations within the DAM.
2. Automated facial recognition
Even more so than with speech, automated face recognition is an essential tool in the armoury of enterprise video management. Especially in situations where large numbers of individuals are involved, automated face recognition and tagging can save thousands of man-hours, as well as ensure that even short clips are correctly tagged.
Applying AI to sporting media
Using AI for sports can automate key functions that include:
● Identifying players and coaches
● Reporting on sponsor coverage
● Integrating player biogs and stats
● Isolating highlights
● Tracking people and objects
● Transcribing game commentary
For more information about how AI can automate sport, media and enterprise workflows, download Imagen’s AI Services Factsheet.
This can be particularly vital to uncover key moments within larger bodies of video, automating an otherwise lengthy process. Appending biographical data can be an extremely valuable resource, for example in a video of a sporting event, where ‘previous games played’, ‘catches made’, and other stats can be turned into a whole new data-based content stream for subscribers or partners.
3. Sentiment analysis and speaker identification
Similarly, speaker identification is essential for audio-only files, whether gleaned from presentations or conference calls. Event video can also be usefully mined for speaker data, bringing in the speaker’s biographical and/or employment data (in the case of a corporate video presentation), or career successes in the case of artists or sports stars, for example.
4. Smart object recognition and tracking
Identifying and labelling everyday objects in video footage can be useful, especially paired with object tracking functionality. This – when powered by AI – can track and tag the movement of multiple objects. The result is that content owners can search for digital media assets more efficiently.
5. Automatic logo recognition
The value of logo tracking from a marketing function perspective, is considerable. But before automation, it wasn’t always practical from a workflow perspective. While relatively static logo tagging (for example, during a pre-prepared presentation) presents fewer challenges, more dynamic and complex environments create a substantial overhead for manual processing.
AI tools however can track relatively fleeting logo visibility in an otherwise lengthy clip – for example, a static sponsor logo in the stands at a football match, or trackside in motorsport. These moments are likely to add considerable value for a sponsor, making them a necessity for larger rights holders and events – a necessity that can now be largely AI-automated.
6. Automatic shot detection
Automatically detecting scene breaks is a crucial element in video management, but it presents more of a challenge than many might expect. However, with AI trained to recognise cuts, fades, dissolves, and camera motion, shot detection is finally within reach for all sizes of business.
7. Automated tag management
The ability to use AI to manage tags is an enormous benefit to any organisation with an extensive library of media assets. From legacy tagging systems and auto-generated tags to accidental manual duplicates, there is a universe of simple tweaks that a good AI tool will immediately analyse and present for approval. Once tags have been culled, edited and managed appropriately, entirely new content opportunities can be uncovered thanks to fast and accurate semantic search across large digital media libraries using simple keywords.
Effective tag management is an investment in operational and content transparency that will have small efficiency impacts everyday, as users easily find what they were looking for, and occasional significant breakthroughs where new, relevant content is discovered at precisely the right time to have maximum impact.
Embracing AI-powered digital asset management
Overall, AI can enrich your media library in multiple ways, most importantly saving teams time and effort, but also by reducing human error into the bargain. By enabling the addition of more varied metadata, and by pulling in other data sources (such as biographical data for individuals in clips), AI goes well beyond manual processing techniques and will add significant value to any business DAM deployment.