You’ll want Nexar’s newly released Live Map for your city

Extracting and exposing valuable insights to enable smart cities and many other applications.

I recently had the privilege of getting a preview of Nexar’s Live Map, from my friend, Nexar’s CTO and co-founder Bruno Fernandez-Ruiz. Nexar uses off-the-shelf smartphones and dash-cams, sophisticated data ingestion, data processing, sensor fusion, and machine learning software to realize their vision of creating the largest safe driving network. To date the company has recorded many miles of driving video (“500 million miles, and more than 40 million miles every week”). This means the company now has access to massive amounts of data (raw video) that has the potential to assist many different constituencies.

We already have access to many mapping services and Google street view, so what differentiates Nexar’s Live Map? Live Map is a free, online, interactive digital map from billions of real-time images and metadata from cameras in the Nexar network. Nexar previously outlined its commitment to protecting user privacy and sketched out its “ethical operating principles”. To this end, data in Live Map is anonymized and aggregated: Bruno and his team have invested a lot into developing technology to mask personal identifiers (license plates, faces, etc.) and other anonymization techniques.

What does Live Map look like? Nexar uses Uber’s H3 hierarchical index to partition the world into hexagons:

Figure 1: Nexar Live Map uses the H3 grid system.

As you can see above, the raw video data is just a starting point. Bruno and his team developed tools to unearth valuable insights (construction zones, traffic signs, etc.) which are made visible through the Live Map interface. The H3 grid system lets users zoom in and out as needed:

Figure 2: Nexar Live Map users can take advantage of time filtering.

Figure 3: A variety of object detection filters are available on Nexar Live Map.

Given his background, Bruno likens the underlying data analysis to what search engines do: they first crawl the web and grab all public web pages, then they use the raw web pages to glean further insights (“derived” or “processed” data) such as URLs/links, entities, images, etc. As Bruno reminded me, very often raw data is just a starting point. This is an important observation for those who have accumulated troves of data: unlocking value often requires some additional creativity and experimentation. In many instances post-processing, refinement, transformation, and information extraction will be needed, before one can feed data into an application or into a model.

After you stare at the images above you can imagine many possible applications for Live Maps, here are two simple ones:

• consumers can use Live Map to view and monitor a neighborhood (24×7) before they decide to move there
• a dashboard for city planners who need to make decisions on a variety of things (e.g., which potholes to fix and in what order)

Nexar’s Live Map is a real-world example of many technologies we are all beginning to talk about: IoT, edge computing, specialized hardware, sensor fusion, real-time data processing and ingestion. Moreover it is an example of an application that seamlessly combines many different machine learning models in the course of an impressive end-to-end system. I’m happy they will have even stronger coverage of San Francisco in the near future!

Images courtesy of Bruno Fernandez-Ruiz.

Related content:

• Bruno Fernandez-Ruiz discusses the tradeoffs we make to ensure safer transportation
• Bruno Fernandez-Ruiz: “Vehicle-to-vehicle communication networks can help fuel smart cities”
• Andrew Feldman: “Specialized hardware for deep learning will unleash innovation”
• “The quest for high-quality data”
• “RISELab’s AutoPandas hints at automation tech that will change the nature of software development”