Four Video Analytics Trends for 2024

At the end of each year, Scylla looks ahead to identify the key trends for the next year. With respect to long-term trends, 2023 has seen ongoing consolidation in the video analytics market, along with a growing shift towards capable cloud-based VMS platforms. An exciting development that emerged this year is the introduction of Large Language Models (LLMs), a disruptive technology that holds immense potential for video analytics. Just like deep learning revolutionized the field in the past, LLMs are set to play a significant role in shaping the future of video analytics. We expect that LLMs will underpin new reporting functionality for video analytics.

1. Market consolidation

The video analytics market went through a “Cambrian explosion” a few years ago, as deep learning transformed computer vision and many startups tried to apply this to video analytics, generally using limited “off the shelf” AI technology.

The market has now matured and consolidated. There are a few vendors with a comprehensive product vision that are thriving, but many are continuing to find it difficult to attain product-market fit and therefore profitability. Some vendors have tried to increase profitability by increasing their prices and reducing investment in development. As a result, innovation has stalled, except for a few vendors.

Since it is now more difficult to raise venture capital for video analytics, more companies that raised capital in 2018-2019 will exhaust their funds and exit the market in 2024.

In addition, solutions purchased in 2018 will be coming to end of life, and buyers will take the opportunity to review the market and potentially upgrade to a solution that better meets their needs. Video analytics buyers will favor vendors that have established product-market fit that offer a complete range of solutions.

2. Move to cloud-based analytics VMS platforms

Traditionally, analytics solutions were deployed on one of a few specialist VMS platforms that dominated the market. These were typically installed on-premises. Now customers are moving to cloud platforms supplied by the analytics vendor.

Three factors have contributed to this: ● Growing awareness and acceptance of cloud deployment as security concerns have been allayed. ● Realization of cost benefits of cloud vs. on-premises solutions. ● Improved cloud VMS options provided by analytics vendors.

Cloud VMS platforms offer the following benefits: ● Analytics can take advantage of the compute resources of the cloud to run analytics as required without needing to scale up on premises hardware for peak loads. ● No upfront capital cost for hardware and reduced total costs due to economies of scale. ● Easy and efficient management of software configuration and straightforward federation for multi-site clients. ● Able to combine edge processing with more comprehensive and insightful data analytics in the cloud.

3. Increasing AI awareness, but applications are lagging

OpenAI’s ChatGPT has had the fastest-growing user base of any application in history, and as a result, awareness of AI is growing. However, the pace of innovation in video analytics has lagged recently. As noted earlier, many video analytics vendors simply applied existing research without significant innovation, so their solutions have limited capabilities, mainly image detection.

Since it is more difficult to raise funding for video analytics now, many vendors have reduced investment in software enhancements.

Some vendors have continued to innovate. For example, Scylla has open-sourced a new loss function, which is now part of YOLO, and has developed new technology that has made its way into a range of new products that can analyze and classify complex motions and which can solve problems in multiple industries and use cases.

Increasing awareness of AI and first-generation AI reaching the end of life is likely to result in customer demand for more sophisticated AI from vendors that can supply this technology.

4. Large Language Models (LLMs) to enhance reporting

LLMs can generate text or images in response to a user prompt based on patterns learned from review of large amounts of text.

ChatGPT is a well-known example of a LLM that can answer questions, perform complex reasoning and even write software.

The potential of LLMs is still being explored but one potential high-value application is reporting for video analytics. A LLM combined with AI-powered video analytics can be used to review analytics, automatically identify patterns, answer questions in natural language, or even respond to verbal requests.

This increases the customer value from AI-based video analytics and introduces a new user interface paradigm.

Conclusion

2024 will see long-term trends towards market consolidation and cloud VMS solutions, with a new trend from disruptive LLM technology being applied to video analytics reporting. Buyers of video analytics solutions should continue to prioritize select vendors that have:

● Their own unique technology that offers benefits over generic “off the shelf” code. ● A comprehensive product line that offers a range of analytics that address a range of real-world security needs (as opposed to a “one trick pony” product). ● Demonstrated capability to continue to develop and evolve their products to benefit from technology improvements. ● A feasible roadmap that includes an integrated LLM to improve reporting of video analytics events.