Ancient Rome didn’t fall for any one specific reason; instead a host of factors including security failures, governmental mismanagement and infrastructure breakdowns all contributed to the decline and demise of the so-called Eternal City.
Similarly, many of today’s cities are facing a multitude of onslaughts, specifically rising crime, acute overcrowding, aging infrastructure, and maximally overstrained services.
Modern metropolises are seeking to avoid Rome’s fate by adopting the philosophy of the “safe city” while implementing new technology-based solutions designed to improve security. Safe cities can now counter the multi-pronged assault on civil life by mustering many different public-safety technologies into a single system that makes the best use of their assets.
Driven by the demonstrated effectiveness of this approach, the global market for safe-city hardware, software and services is expected to grow to more than $20 billion in 2020, up from $13 billion in 2016, according to a forecast from IHS Markit.
Municipalities recently gained a new set of tools to build effective safe-city systems, including emerging artificial intelligence software and services. Ideally, cities could deploy a unified AI platform that leverages the collective power of many specialized cognitive engines to process many types of information gathered from many sources. The Veritone Platform represents such a structure, one that is specifically designed to orchestrate cognitive engines, make them work in concert and empower them to produce actionable intelligence.
Using the Veritone Platform, metropolises can efficiently and effectively utilize resources including video surveillance and biometrics. This allows governments to improve the security of citizens, businesses, and visitors—and truly attain the vision of the safe city.
The world is undergoing an explosion of urbanization. About 100 years ago, only two out of every 10 people on earth lived in an urban area, according to a World Bank estimate. By 2008, half of the world’s population lived in cities. This trend will continue in the future, with cities housing 70 percent of people by 2050.
At the same time, many cities are facing an increase in illegal activity, with the FBI reporting a rise in homicides, robberies, aggravated assaults and non-fatal shootings in major U.S. metropolises in the first half of 2016 compared to the same period in 2015.
These factors are contributing to an increase in stress on urban infrastructure and public-safety services, including police, fire and emergency medical.
To manage these issues, cities around the world have made massive investments in new technology. Government-controlled IT systems now collect all kinds of structured and unstructured data, including video, audio, passport control, traffic pattern, hotel registration, GPS feed, access control and building security system information.
The amount of data produced by surveillance systems alone is astounding, with new security cameras expected globally to generate 815 petabytes of data every day in 2019. By the end of 2016, about 350 million surveillance cameras had been installed worldwide, IHS Markit estimates. It’s simply not possible to manage this massive quantity of disparate data using traditional techniques of manually poring over hours and hours of content.
Surveillance and tracking technologies by themselves don’t make a city safe. A method must exist to amalgamate the information collected from these disparate sources into a single IT platform. Only then can a technology platform deliver a comprehensive picture of urban public safety and facilitate communication and collaboration among multiple law enforcement, prosecutorial, and public safety agencies.
Luckily, a large and growing number of AI cognitive engines are now available to perform the task of processing structured and unstructured data quickly, cheaply, and with a high degree of accuracy. But how can municipalities best identify and use these engines to their advantage? And, as importantly, how can they future proof their investments against heterogeneous technology and as cognitive engines evolve?
Relevant Cognitive Engines
Veritone employs best-in-class artificial intelligence engines in a variety of cognitive classes, uniquely activated to produce actionable intelligence either on a near real-time or evidentiary basis after the fact.
- Transcription engines utilize natural language processing (NLP) to analyze audio streams from security cameras body and dash cameras, and interrogation footage to produce transcriptions in all major languages. When coupled with speaker separation, text can be segregated by conversation participant. Some engines can be trained for very specific use cases such as the codified vernacular of criminal gangs.
- Translation engines translate written text from one language to another and use algorithms to increase the accuracy of sentence structure and parts of speech.
- Face detection engines identify the presence of an anonymous face in video, and correlate that face across multiple video sources. Face detection can be augmented for facial identification by utilizing a reference library of known people, and can also be extended to the redaction of anonymous faces other than a key subject. Some face engines detect and describe emotions, such as anger or fear.
- Object identification, also known as computer vision, engines are trained to identify either general or specific classes of objects. In the public safety setting, object identification can be used for the make and model of vehicles, identification of weapons, unattended luggage, or other specialized tasks.
- Audio/video fingerprinting engines identify segments of media that are a match to certain sound profiles. These engines use specialized libraries of reference audio clips, such as gunshots, sirens, or specific phrases used in law enforcement.
- Landmark identification engines can be trained to recognize certain landmarks as a means of triangulating activity from non-stationary cameras, such as video footage taken by mobile phone.
- Geolocation engines compile and track coordinates and time synchronization across multiple video sources such as surveillance cameras.
The Veritone Platform is specifically architected to be highly configurable and intuitive, providing the optimal artificial intelligence solution for metropolises that want to harness the power of AI in their safe-city initiatives. It is a continuously-evolving environment of valuable applications, core services, developer tools, and cognitive engines.
By combining superior cognitive capabilities in a single platform, Veritone removes the need for governments and agencies to choose individual vendors in each cognitive class from the landscape of more than 5,000 engine developers, thus providing a solution that will continually improve with multiple use cases.
Private and Shared Libraries
Embedded in the Veritone Platform is the capability to maintain secure private libraries of tagged images for facial recognition, license plate recognition, vehicle recognition and landmark recognition specific to any public-safety organization’s needs. Libraries can be pre-populated from local databases, such as repositories of known criminal suspects, or they can grow organically over time based on the tagging of individuals or objects from within Veritone’s intuitive user interface.
Veritone libraries can be operated privately or shared across government agencies at the local, state and federal level, with complete control and visibility of user accounts and activity.
The Veritone Platform will offer a variety of implementation and media storage options to match the technical environment and security requirements of clients. Current deployment options include Amazon Web Services (AWS) and Microsoft Azure Government in the United States. Additional deployment options for AWS in the United Kingdom, AWS Government in the United States and a hybrid on-premise cloud version of the award-winning solution are in development and scheduled for release in the second half of 2017.
The Veritone Platform ensures data privacy while affording the ability to share with other parties quickly, efficiently and securely, independent of deployment model. Veritone also works with its clients to optimize media storage options based on client budget and archiving needs. The Veritone Platform offers several options, such as:
- Immediately purging media assets after processing, saving only the resulting metadata.
- Temporarily storing media assets in the cloud on local hardware, with migration to periodic long-term backup storage.
- Processing of media assets already residing in client’s long-term storage stack via direct read.
Operates with Any Installed System
Veritone easily facilitates media ingestion via a wide variety of proprietary and open protocols, including security and law-enforcement camera and media management systems. The Veritone Platform technology stack is built to be plug-and-play with external media sources and can be customized for virtually any media origination scenario.
Furthermore, the Veritone Platform system can work with older surveillance technologies, adding AI capabilities without requiring investments in new cameras or other devices.
The Veritone open developer ecosystem is designed to attract and retain cognitive engine developers. The model will allow for easy onboarding, configuration, testing and competitive benchmarking across one or many of our deployment models.
The Veritone Platform is designed to promote and maintain a mutually beneficial commercial relationship with engine providers by directing processing volume, and thus monetization opportunity, to the best narrow-intelligence engine based on client requirements for optimal cost, speed and accuracy.
To make the information processed by the Veritone Platform useful, the platform supports third-party applications and includes powerful native applications such as:
- Veritone CMS, which ingests and stores media within a robust, searchable library and makes it readily available for analysis and usage.
- Veritone Discovery, which dissects and analyzes media with powerful indexing and multivariate search capabilities.
- Veritone Collections, which efficiently and provisionally shares, embeds and distributes content.
- Veritone Admin, which is used to establish, monitor, and manage account configurations such as user configurations and cognitive workflows.
The Veritone Platform is easily configurable to create custom, rule-based workflows incorporating cognitive processing in parallel as well as in a sequenced or chained fashion. A simple example:
- Video camera feeds can be set to automatically process for license plate identification, and when found those Clips can be saved into a Watchlist in the platform.
- Once saved to Watchlist, the license plate number can be appended to secure, structured vehicle registration databases to retrieve pertinent vehicle information.
- Key video frames can simultaneously be routed to an object identification engine in Veritone CMS to identify the make and model of the vehicle.
- Where there is a mismatch, potentially indicating a stolen vehicle, the reference clip can be shared to a pre-selected group of law enforcement personnel as an alert using Veritone Collections.
- Video clips with mismatching registrations and recognized vehicles can later be searched via license plate number and/or vehicle make and model in Veritone Discovery.
“Video content analysis is becoming more important, especially as cities increase their installed base of cameras…there’s no way just a couple of people can analyze all that data, so it’s extremely important to employ video content analysis.” Alexander Richardson – Critical Communications Analyst, IHS Markit
All Roads Lead to Rome
With 21st century cities under siege from a multitude of challenges, public-safety institutions need to use AI to get the best results out of their technologies and attain the vision of the safe city. Veritone represents an innovative new approach to this problem, intelligently orchestrating today’s confusing plethora of cognitive engines to transform the exploding amount of collected data into useful and actionable information that can tangibly improve urban safety.
One can only imagine that if the ancient Romans had access to such technology, we all might still be speaking Latin today.
John Newsom is executive vice president at Veritone. He is a software executive with an evangelical passion for AI technology who aligns the Veritone Platform with customer and market needs.