Security Leader Spotlight: Qualcomm’s Siddhartha Franco

Periodically in this space, ISC News Editor-in-Chief D.J. Murphy will sit down with a leader in the security industry who will share their thoughts and expertise on issues affecting a variety of industry stakeholders. This month we feature Siddhartha Franco, director of business development for Qualcomm. Well known as a leader in wireless technology, Qualcomm has a growing IoT business supplying technology for connected cameras for many applications including security. Franco defines the “connected intelligent edge” and discusses how surveillance technology has integrated AI and machine learning to get there.

ISC News: What is “the connected intelligent edge” and what does it mean relative to security, surveillance and monitoring?

Siddhartha Franco: In general, edge architecture is a new paradigm that brings computation and data storage closer to the sources of data for improved response times and efficiency. The effectiveness of a security system relies on how much data it can process in the service of producing real-time decisions that make buildings, property and people more secure, with the end goal of affecting people's lives in a positive way. Processing data closest to the edge provides a private and secure way to make real-time decisions.

At Qualcomm Technologies, we are making edge AI ubiquitous and enabling the connected intelligent edge by researching and developing power- and performance-optimized machine learning solutions. We are building on the scale, rapid innovation, and power efficiency of mobile by expanding AI to other devices, machines, vehicles, the cloud, and things. The connected intelligent edge powered by a powerful SOC (System on Chip) allows for running multiple complex AI models on multiple input streams from cameras as well as alarm sensors. Moreover, 5G provides a means to connect the unconnected, reducing costs of cabling while enabling high bandwidth communication between connected spaces as well as the cloud

ISC News: How does operating security departments change after implementing smart connected solutions?

Franco: Traditional systems had security personnel staring at a giant wall of monitors—a largely manual process prone to errors. Moreover these systems were installed one time with a fixed configuration. Recently they have been running analytics, but that too is limited to the time of installation with no easy way to update or add newer use cases. And then you have all the IT requirements for servers and storage on site.

With video software-as-a-service models, enterprises are now able to deploy plug-and-play cameras that can be configured and managed remotely. With state-of-the-art connectivity, you can easily get a live stream from any camera from anywhere in the world. The capability to run machine learning models on the edge, power efficiently allows you to bring new AI based use cases to the edge. You can also easily upgrade or remove AI features as you need and within the same network have different cameras behave differently (e.g., you want to detect masks at the entrance of a lobby, but elsewhere you want to people-counting to detect a crowd). You can run different AI models and orchestrate them to your needs. The connected intelligent edge enables this fundamental shift with numerous applications in the security industry.

ISC News: How is Qualcomm making it easier to implement these kinds of systems?

Franco: Qualcomm is committed to continuously building a broad portfolio of SOCs to power smart camera and edge boxes. We’ve also created a very broad ecosystem of leading ODMs and SOM makers that work closely to fully utilize the power of Qualcomm processors for our end customers.

As more and more problems are being solved with AI, we continue to broaden the ecosystem of AI partners by simplifying AI development through our rich set of tools and AI SDK. 

• Qualcomm Neural Processing Engine: AI conversion tools to solve the common pain point to convert server trained AI for edge inference 
• AIMET (AI Model Efficiency Toolkit): Tools and model zoo to enable ecosystem partners
• PTQ/QAT (Post Training Quantization/Quantization Aware Training): Continuous pursuit in AI model efficiency 

Working with others in the industry, these technologies can be implemented into an end-to-end solution for the industry to deploy and commercialize as a VMaaS that ultimately helps to build safer and more protected environments.

ISC News: What are some examples you have seen of a security department adopting these kinds of systems?

Franco:  Flock Safety is a customer based in Atlanta that provides automated license plate recognition technology to law enforcement agencies. It has deployed both our camera and AI Box solutions with an ambitious goal of reducing crime in America by 25 percent.

Using Qualcomm’s Vision Intelligence Platform-based cameras, they developed a location-flexible, easy-to-install advanced license plate reader that goes beyond license plate recognition to detect vehicles through means of other characteristics like audio, sound and geography. Working with Qualcomm and Thundercomm, Flock Safety was able to integrate and implement both solutions including cellular connectivity and quickly go to market with new products enabling LPR, like their Falcon Flex LPR cameras which uses on-device ML, and sends the most important data to the cloud for real-time analytics. The result is not only cost effective and easy to deploy with the ability to utilize existing infrastructure, but offers real value to their customers like law enforcement agencies to solve and eliminate crime. One camera deployment in Wichita, Kan. resulted in the recovery of 11 stolen vehicles worth nearly half a million dollars and 49 arrests in the first 30 days.