Vice President, AI/ML Products
The initial euphoria over AL/ML seems to have died down in the M&E industry. Our research shows that many M&E players have run AI initiatives with different vendors but have not achieved anything substantial enough to solve their business problems. Though the demo was impressive, the project hit a wall at the Proof of Concept (PoC) stage because the AI solution did not work for their content! When the cycle was repeated with multiple vendors, they concluded that AI models are not available or mature enough to solve specific M&E business challenges.
The fact is, no M&E client’s business problems in their entirety can be solved effectively by any one AI engine or solution provider in the market. Also, the heavy lifting involving trial and evaluation of multiple vendors rests on the client’s overburdened shoulders. But do they have the required data science talent in-house to tweak the AI/ML engine for their enterprise’s data model and ensure accuracy and actionability? The lack of adequate expertise also scuttles any in-house project which attempts to build an AI/ML model.
To crack the impasse, what is needed is a media recognition AI/ML platform that brings the best-of-breed AI models and home-grown niche models to address the issues of accuracy and actionability. Plus, to tackle the talent gaps, consulting expertise in deep learning AI with computer vision knowledge is critical. For AI/ML to deliver for M&E organizations, no cookie-cutter approach will work; what is required is a tailored, bespoke model that embraces the unique data nuances of the client enterprise.
At Prime Focus Technologies (PFT), we believe that if AI is to work for M&E players, then it has to deliver accurate and actionable data which can solve unique business challenges. For this to happen, solution providers have to be open and committed to work with any AI/ML engine, have the data science talent pool to interpret the data and its subtle nuances, and tweak it to suit the needs of the enterprise’s content.
While the expectation from AI right now is to solve for accuracy alone, PFT has gone a step further and tried to solve for actionability as well. We offer a combination of Technology and Consulting to deliver accurate and actionable data that can solve specific M&E challenges seamlessly. This is how we make AI work for you!
PFT’s native media recognition AI platform CLEAR Vision Cloud helps solve real world business problems of TV Networks, Studios and OTT platforms because of its perfect combination of both technology and consulting. It integrates best-of-breed AI engines like Microsoft, Google, Amazon Web Services, IBM Watson, and home-grown AI models along with a unique Machine Wisdom layer that is focused on harnessing the best quality data. Along with Technology, PFT’s bespoke strategic Consulting services ensure AI works for the customer, taking into consideration their specific business challenges and unique content.
CLEAR Vision Cloud offers AI data at 3 levels (Basic, Advanced, Premium), along with additional data packs including Compliance, Comparison and Transcript. It also provides a set of powerful Action Toolkits that makes the AI harvested data actionable. Action Toolkits, as the name suggests, are ready-to-use to address specific M&E use cases. The Action Toolkits include – Discovery, Segmentation, Video Comparator, Content Moderator and Language Tools. These are enveloped by PFT’s unique Machine Wisdom layer that imparts cognitive benefits to CLEAR Vision Cloud. Think of it as an AI platform with a human brain!
CLEAR Vision Cloud AI Data Pack for Basic Metadata helps identify physical video segments (blacks, color bars, slates, pre-caps, re-caps, montages, essence); text & textless segments; specific captioned segments; and custom segments based on customer need across long form and short form content with 100% accuracy & 100% frame accuracy.
The Segmentation Action toolkit allows review & QC automatically identified segments; filter out content segments & export EDLs; and generate a video of the custom segment by stripping out the rest of the physical segments. Substantial reduction of time and cost of
segment marking, and content segment extraction is achieved. Automatic learning of segment signatures based on QC input is an industry first and a key enabler in workflow automation. While we have built a basic toolkit for segmentation, there are many requirements that can be addressed and many use cases one can think of. We believe AI has the capability to address all of these effectively. PFT’s AI model is home grown and customized to solve specific M&E use cases to make AI work for you.
For example, a leading TV Station Group in the US was manually identifying and marking segments like color bars, blacks, slates, and content segments in the ads they receive for broadcasting in their network. This dependency on operators was time-consuming and adding workforce costs to them. There was always a possibility of error due to manual intervention.
Though there are a few technology solutions available, they have limited capabilities and cannot handle the vagaries and noise in the content. Solutions that identify blacks using an image or signal processing cannot tolerate black and color bar variations from various sources. Detection of more sophisticated segments like slates needs deeper cognition and interpretation of the content, missing in the available tools.
The Station Group needed a solution that could precisely, and frame accurately identify these segments every time to ensure complete automation of the workflow. PFT provided its powerful AI platform, Vision Cloud combined with Consulting Services to address these challenges. Vision Cloud is a sophisticated AI platform that recognizes these segments precisely as well as frame accurately in the ad spots. Vision Cloud employs several computer vision and deep learning technologies trained with huge amount of past data to natively develop this solution. Machine Wisdom, PFT’s patented technology is used to make sense of the cognition delivered by underlying AI & ML models. As a result, Vision Cloud can identify and detect color bars, slates, and black screens accurately, even if they are from different sources and come with variations and noise. After performing a QC, the spots are sent for playout.
PFT’s Vision Cloud has brought the power of AI to the customer and ushered in complete automation of the workflow with no manual intervention. The effort and cost saved as a result of this automation can be invested towards creative pursuits aimed at making the content more compelling and engaging.