The Unleash live platform is a powerful tool that allows A.I. computer vision models to be applied to live camera feeds.
Unleash live runs high-performance analysis on HD video streams at up to 30 fps in real-time, giving cities, infrastructure owners, and public safety unparalleled insight into their operations.
Not only does Unleash live run A.I. inference, but all the A.I. generated analytics are live-streamed, which may be visualized to extract valuable insights.
Our customers build fantastic custom insights from powerful camera analytics. These insights span many critical use cases, such as:
-
crowd movement,
-
fruin analysis,
-
traffic flow,
-
traffic congestion,
-
zone utilization,
-
parking utilization,
-
curbside management,
-
queue analysis,
-
multi-modal transport analysis,
-
security zone breach analysis
-
…and many more.
Esri’s ArcGIS Online offers state-of-the-art dashboards to display such data in real-time.
This blog post shares a high-level overview of how we integrated Unleash live A.I. Analytics with Esri Dashboards and how it improves our customer’s decision-making. If you want to learn more about a real-world problem we solved, read more about our partnership with the City of Sacramento, California, USA.
The raw A.I. data challenge
Unleash live can stream with 200 A.I. inferences per second. It is a tremendous firehose of information. Our customers, however, often request a consolidated custom insight to identify trends or pick an insight over time.
So, we are post-processing all the A.I. data in a serverless and highly efficient cloud computing setup. Once cleaned and piped through an API, Esri gets accurate and meaningful data for insightful presentations.
Esri’s ArcGIS Online Dashboards
Many of our customers already use Esri ArcGIS Online and are not looking to adjust their insights gathering and presentation workflows. Unleash live fits right into this existing workflow.
One functionality our customers, in particular, appreciate is the ability to show any type of data on Esri dashboards. The default workflow is simple — prepare a CSV file on Unleash live, upload it to the Esri platform, and then build a custom dashboard based on the imported data. This works for users who need to visualize their data in a static form, and indeed, our very first MVP was built this way — the challenge was to bring live data to dashboards, unleashing a whole new world of valuable use cases.
ArcGIS Online is fed from a REST API, so we sat down with Esri and discussed how to feed newly generated live data into the system easily. This sounded easy, especially because the Esri API has a Python wrapper!
Edge conditions
Unleash live heavily utilizes AWS serverless functionality to scale horizontally with the number of events handled. Additionally, we need to keep track of execution time and the number of invocations since these directly translate to the cost of our solution.
To reduce the number of invocations, we decided to queue detections and send them to AWS in batches. This was easily done since Python ships with a threading module, which is extremely simple to use and extend.
Integration
We took an incremental approach to reach minimal viable functionality quickly before improving it incrementally. The first iteration of this feature simply automated what users would do to generate the CSV and upload it through the API.
We quickly realized that this implementation can take up to 35 seconds per API call. We expected the execution time to be much less!
It turned out to be a minor problem with the Esri API server. Through our close partnership with Esri, we quickly solved this challenge and were also able to reduce the number of expensive API calls. The result was an adjustment to our solution with better-suited API calls. This allowed us to bring down execution time to 8 seconds. These 8-second executions worked well with flushing the queue every 10 seconds.
Dashboarding
Once we had the highly efficient API architecture setup, we built powerful Esri dashboards. Once fed with live A.I. data from Unleash live, Esri allows users to visualize data from multiple sources concurrently. Our goal was to deliver a last-hour snapshot, which made our dashboard almost live — almost because Esri will not refresh automatically (yet) — however, we are now working with Esri on bringing this new feature improvement into the platform. Stay tuned!
Conclusion
Expressing data on dashboards delivers powerful insights to end-users. Data that looks dull in CSV comes to life when visualized live, often unveiling patterns or trends that would otherwise be difficult to spot. At Unleash live, we recognize the value of data visualization and bring it to the next level by feeding dashboards with data from AI-powered video analytics.
In the above example, we used a smartphone’s camera with an overlaid A.I. model, which we set up on Unleash live in seconds. It took us longer to position the phone than to set everything up on the platform!
Get Started with Unleash live
Visit unleashlive.com to learn more, or book a time with one of our specialist team members to learn how your business can take advantage of live-streaming video and A.I. analytics.
