Artificial Intelligence & Video Analytics: Why is it a hot topic, and what is it's future?

AI field has become a ubiquitous part of our daily lives, from hailing a cab to receiving a simple weather prediction for the day, to complex task such as stock market prediction or personalized movie recommendations on Netflix. AI is so deeply embedded in our routine that we may not even realize it’s presence.

The expansion of AI is fueled by various factors, including the availability of data, affordable computing and connectivity infrastructure, government funding for research, advancements in semiconductor technology , the availability of data science talent and the catalyting impact of Covid-19 pandemic. These factors have combined to create an environment in which AI growth has accelerated and become pervasive in our lives

The emergence of new startups and collaborations is one of the many outcomes of the field of AI, which, like the internet and computers, is poised to transform various industries and provide solutions to wide range of problems. AI is transforming the way we think and work, and is making a profound impact on our society. If I may say that “AI represents the biggest revolution since the dotcom and smartphone paradigms” , I may not be far from the truth.

AI based Video Analytics:

Today, a significant portion of available data or information for both humans and machines is in the form of videos. According to Sandvine on Cision PR, video usage grew by 24% in 2022, making it 65% of all internet traffic.

Video analytics (VA) is an area of AI that has made significant strides in recent years. Various machine learning algorithms can process video data streams from IP cameras in real-time, performing multiple parallel tasks, such as image recognition, object detection, classification, anomaly identification, and?predictions. This ability to extract valuable analytics, metadata, and alerts from traditional IP cameras through VA software has transformed them into "smart" cameras that aid in decision-making. IP cameras with pre-installed VA software, categorized?as?AI-cameras have come into existence.

VA offers a variety of analytics, such as face recognition, people counting, automated number plate recognition (ANPR), fire detection, fall detection, intrusion detection, and other capabilities. By providing real-time insights and enabling proactive measures to prevent incidents before they happen, this technology has the potential to transform security?and?safety.

Industries & applications using AI Analytics:

The utilization of AI analytics is propelled by both societal and commercial necessities, which encompass safety, security, efficiency, revenue generation, compliance, and cost reduction. A few applications?are?as?follows:

• AI technology can be used in elderly care facilities to detect falls and notify caregivers for prompt assistance.
• In a Smart Parking system, AI can identify instances of illegal parking.
• In Smart Retail, AI can monitor customer behavior and preferences to offer customized recommendations or to recognize the presence?of?VIP?clients

Industries or sectors that experience manpower shortages or require improvements in safety and monitoring adherence can benefit from AI technology. For instance, AI can be employed in factories and construction sites to supervise adherence to safety regulations and confirm that workers wear proper protective gear. Similarly, AI can be used in hospitals to supervise compliance with dress codes and restricted areas. In general, the utilization of AI analytics has the capacity to revolutionize industries and enhance daily life across numerous?domains

The Emergence of edgeAI:

In the earlier days, obtaining analytics through video processing was mainly carried out in the cloud as there was limited computing power and high implementation costs at the "Last Mile" or the "Edge." However, with technological advancements in VLSI geometries beyond 7nm, video processing is now shifting from the cloud to the edge-computer, which is commonly referred to as an?"edge-AI?box."

To provide context, the term "the edge" is defined by Accenture as processing data in closer proximity to its point of origin, resulting in faster and more substantial real-time outcomes due to greater speeds and volumes of data- ?commonly referred to as?edge?computing

By processing data at the edge instead of transferring it to the cloud, businesses can lower latency, decrease data transport expenses, and enhance their ability to make real-time decisions. This is especially beneficial for applications that necessitate immediate action, such as industrial automation, real-time surveillance, and autonomous?vehicles

EdgeAI enhances the value of CCTV Infrastructure:

Low-cost CCTV cameras are now widely available and are being used for more than just security and surveillance applications. However, their usefulness can be greatly enhanced by adding edge-AI boxes with video analytics software to them. Today, there is an increasing demand for Smart Cameras that can offer analytics beyond traditional IP Camera for only-Security/Surveillance use cases, which are becoming less popular. The task of manually monitoring and interpreting large amounts of data, whether visual streams or other datasets, is challenging. Without the assistance of AI, it becomes even more difficult to visualize and gain actionable insights on?a?large?scale

By integrating with a Video Management System, edge-AI can facilitate notifications and recommend actions based on pre-established rules, resulting in better decision-making at the edge. This not only boosts security but also offers novel perspectives and business advantages for the user. Edge-AI is also gaining traction in mobile CCTV surveillance deployments, which is a burgeoning market with enormous?potential.

The proliferation of edge-AI can be attributed to various factors including:

• Heightened privacy and security concerns associated with the vulnerability of video feeds to hacking, which necessitates keeping them within the premises.
• Increased flexibility in designing the edge network or architecture.
• More rapid decision-making and action at the edge.
• Decreased data transportation to the Command Center
• Edge allows faster processing and removes the need to stream videos onto the cloud. It can balance out heavy loads that would otherwise result in a lot of cloud costs too.
• A decentralized or distributed architecture provides resilience, reducing the risk of a single point of failure (such as failure at the Command Center)
• Edge-AI-based systems are affordable, with no recurring or hidden costs, unlike cloud-based systems
• Reduce manpower operational costs

"FPGA" chips: The preferred Edge-AI hardware solution”

FPGA (Field Programmable Gate Array) chips have become a popular choice for edge AI hardware, alongside traditional GPUs (Graphics Processing Units). This is because FPGA chips:

• Offer lower latency
• Consume lower power
• Provide real-time parallel processing
• Are scalable and flexible, allowing for easy code migration from lower to higher variants
• Easily support field upgrades to accommodate ever-changing ML models.

Singapore-based companies like Planetspark are at the forefront of developing innovative, state-of-the-art edge-AI boxes that utilize AMD FPGA chipset. By collaborating with local and overseas algorithm partners and start-ups, they offer unique edge-AI solutions, such as video analytics for various use cases including car parks, theme parks, construction safety, and residential security.?

Additionally, as LiDAR technology continues to mature, the combination of vision and LiDAR will become increasingly popular for e.g. in smart road infrastructure and indoor use cases. As a result, real-time and parallel processing, flexible architecture will be required, making FPGA the ideal solution. (Further discussion on this topic will be included in an upcoming article).

The future of analytics is in the edge-AI:

According to data from Comparitech and HIS Markit cited in a report by Visual Capitalist, Singapore currently has 18 cameras installed per 1000 people. Several cities in India and Vietnam are also among the leaders in cameras per capita globally, excluding China. With such extensive CCTV infrastructure in South Asia, there is significant potential for various industries and use cases.

As 5G and IoT connected devices continue to be deployed and data availability and demand increase, the future of edge-AI analytics looks promising. Along with video feeds, more sensors will be connected, and edge-AI is expected to transform diverse sectors such as logistics, factories, education, construction, agriculture, healthcare, and more. This presents a significant opportunity for Singapore and other South-Asian countries to collaborate with companies like Planetspark and participate in the rapidly evolving edge-AI transformed business world!

Stay tuned ....

Firstly, my Thanks to LinkedIn for facilitating this enterprising platform and Thankyou all, who are part of my lovely LinkedIn community (the readers, my connections) to invest your time in this article. I see this world as a learning and cherishing journey and hence look forward to learn from you too!

If you need the article in pdf with graphics, or look forward to collaboration with us in this AI-edge field, I would be happy to learn and understand from you and share more . Please email me at: [email protected]

And above all, my special Thanks to my wife Beena, my Excelpoint Management and colleagues, AMD team, Planetspark team, SG Tech friends, my customer partners, and Tech company mentors who guided me with the insights?and my learning.

Author: RD Pai , VP Business Development @ Excelpoint - Singapore ([email protected])

About me: I am an Engineer in Electronics & Telecommunications with 30 over years of experience in diverse Tech domains (Semiconductor Components to Embedded to AI & IoT edge) spanning AsiaPac. I am a passionate proponent of digital transformation and EXCO member at SGTech Smart Nation Chapter, a member of SCS IoT Chapter, observer at AEIS and member at IoT Community.