The demand for edge analytics is increasing rapidly with the explosion of streaming data from sensors, cameras, and other sources. Of these, video remains the dominant data source with over a billion cameras deployed globally. Enterprises want to extract intelligence from these data streams to create business value. As such, over the past decade, we’ve seen an explosion of intelligent video analytics (IVA) offerings. Initial attempts at IVA, like motion-detection analytics, were infamous for generating runaway false positives from environmental factors like shadows, weather, and moving lights. Recently, however, AI machine learning- and deep learning-based analytics have addressed many of these weaknesses, while extending capabilities far beyond early IVA technology.
These advances have led organizations to adopt IVA not only for reducing risks for physical security, but also for improving efficiencies in business operations. Many organizations now see extremely compelling ROI that can be achieved through IVA, from both reduced costs and extended capabilities.
Yet, many security, IT, and digital transformation leaders are struggling to achieve the returns promised through a move to an AI-driven organization. Interviewing customers and industry experts over the past two years, we’ve seen several common issues emerge that prevent wider adoption of AI-based video analytics technologies:
- Inconsistent accuracy – Many analytics work well in controlled environments, but seemingly simple use cases are often more complex than anticipated, with resulting impact to accuracy
- Customization and integration for operational reliability – Unique use cases and business requirements require customized models (and, in some cases, integration with custom applications), so that businesses can actually change processes and headcount based on their deployed IVA solutions
- Performance is essential – Businesses are challenged by IVA solutions that often do not dependably provide real-time performance at the lowest TCO
- Scalability problems – Whether it’s scaling out across multiple sites, adding more channels to existing systems, or the need to layer additional analytics to existing IVA systems, businesses need multiple compute and deployment options, without supply chain risk
We believe a new approach—using a platform built on the principles of Open Analytics—effectively addresses these issues. Open Analytics solutions are built upon three pillars:
- Open customization: High levels of control, including end-to-end configuration of the analytics pipeline, ability to swap/retrain/redeploy AI models, and ability to configure or deploy filtering libraries to implement advanced rules
- Open choice: High flexibility through vendor/provider independence, cross-platform compute options (CPU, GPU, FPGA, SOCs), multiple deployment models (edge, on-prem, private/public cloud, and hybrid), and wide data source (video and other streaming data) support
- Open integration: High connectivity, including API(s) to integrate with business or security systems, multiple application and UI options, and SDKs to embed IVA functionality into any application or product
Open Analytics offers a new way forward:
- Instead of canned, out-of-the-box AI models, Open Analytics provides flexibility to adjust the entire AI pipeline (including models and filtering) to achieve better accuracy and performance.
- Instead of vendor lock-in and exposure to supply chain risk, Open Analytics supports multiple compute platforms and vendors, and the ability to easily switch on the fly.
- Instead of hidden costs resulting from overly-optimistic hardware specs, Open Analytics offers transparent, pre-tested requirements, and a full range of price to performance options for best TCO.
With Open Analytics platforms, end users, integrators, and developers can now drive deeper adoption of IVA technology into their organizations to achieve significant ROI from smart building, smart city, smart factory, and smart office initiatives.
Whether you’re an end user or integration partner searching for an intelligent video analytics solution that you can depend on, or a developer or OEM that wants total control over how you embed an AI platform into your application supporting multiple data sources, solutions based on Open Analytics platforms provide the best of both worlds.
You can meet your current security and operational needs with IVA and provide a path forward to integrate streaming data from sensors and other devices into the same Open Analytics-based platform to provide enhanced capabilities.