Smart Cities and Image Recognition

Image recognition and machine vision have real-world implications for local governments and can help officials efficiently integrate and manage assets

Photo from Pubnub.

Originally posted by Joe Hanson in Towards Data Science.

Advances in artificial intelligence mean applications can increasingly take on image recognition capabilities that allow them to identify objects, detect the age of human faces and screen out adult content. The Department of Homeland Security has worked for several years to implement a biometric monitoring system to verify travelers in U.S. airports, and they recently found success with a Customs and Border Protection pilot.

The system uses facial recognition software to compare photos of passengers against a database, allowing DHS officials to identify travelers who have overstayed visas or are wanted in criminal investigations.

These developments underscore the need for the government to remain abreast of ways to manage complex technology and maintain standards of living. Image recognition software has real-world implications for local governments and can help officials efficiently integrate and manage assets.

Gif from Micael Reynaud.

Image Recognition Opens Up a New World of Opportunity

Image recognition software can now process an incredible number of images at unprecedented speeds, all using completely serverless technology — a staple of advanced image recognition APIs.

Innovative image recognition companies like Clarifai or Sightengine are focused solely on building the technology, using machine learning to continue to improve their APIs for accuracy and speed. They’ll handle all the processing and systems. For a developer, it promises and delivers the serverless experience.

Image Recognition + Functions

Moreover, image recognition software can integrate with other APIs through functions to trigger various actions. When a camera senses a certain object, for example, it can fire off a text message or use a content moderation API to filter the image. It can do all this while data is streaming from source to endpoint.

Gif from Micael Reynaud.

This integration potential allows us to build smarter systems. For instance, developers could pair image analysis with a weather application. A combination of actual event data with nearby and related information can add to the value of the event streams, accurately predicting and measuring the effect of recognized patterns on residents and businesses. The best part? Image recognition software becomes more accurate over time.

Better Learning Through Machine Learning

Technology has evolved beyond simple pattern recognition thanks to machine learning. Rather than merely identify similarities and differences among a series of images — like the characters “3,” “8” and “B” — applications can use machine learning to perform more complex analyses.

Computers can learn to draw their own conclusions about images or videos, tagging them accordingly. The more times a computer performs this sort of analysis, the more capable it becomes of correctly identifying and tagging other images in the future. Larger data sets lead to more accurate results, and feedback loops help eliminate errors.

Gif from Micael Reynaud.

Fighting Traffic with Image Recognition

An awesome example of this happening in the real world is happening on the busy streets of New York City. The NYC Department of Transportation has partnered with to use image recognition to better understand major traffic events.

A combination of image recognition and machine learning enables the system to detect traffic jams, weather patterns, parking violations and more, sending real-time alerts to city officials along the way. The intersection is rigged up with cameras that capture activity, process it, and stream back findings and actionable data in realtime (like sending traffic enforcement for illegally parked vehicles, or service for broken or malfunctioning equipment.)

And in the spirit of APIs working together in harmony, as video recognition events occur, the system instantly blends weather, demographic, and location-specific data with the original event data to provide highly embellished realtime feeds that are ready for analytical consumption in dashboards that support realtime analytics.

Check it out in-action below:

This sort of collaboration represents a fraction of what’s possible with image recognition technology. By using the technology to more efficiently address the problems of today, cities can be prepared for the world of tomorrow.

Machine Vision Flourishes Where Unconnected Devices Falter

Machine vision and deep learning have matured to become essential tools that can be applied to expand the boundaries of IoT. Almost all of the physical world we live in is unable to participate as IoT devices. Machine vision, however, stands poised to accelerate the inclusion of real-world activities as IoT data.

IoT devices are only valuable if they’re connected. But what if they aren’t? How do we get data on them if they aren’t connected to the internet? Image recognition. We can analyze traffic, we can analyze sanitation, we can analyze overcrowding. All this takes a robust and powerful image recognition software, that can learn from the patterns, and get the data where it needs to go.

Photo from Pubnub.

Data Streaming

A big piece of building smarter cities with image recognition is the data streaming elements. In other words, how do we deliver the data from source to endpoint in real time? Data streaming infrastructure will play an increasingly important role in this. Even more powerful will be image recognition (and other APIs), within the network, allowing cities to serverlessly deploy their image recognition, run the processing on the network while the data is in motion, and deliver the processed data to its endpoint, whether it be a dashboard or an alert to the relevant person.

Recognizing Untapped Potential

Optimizing real-world infrastructure

Cities will be able to use image recognition to better understand how people and things move. With a wealth of data at their disposal, urban planners and city officials can make more effective decisions about where to improve physical infrastructure for optimal safety and efficiency.

Eventually, applications equipped with image recognition technology will be able to make decisions without humans. Visitors to the recent TechCrunch Disrupt Hackathon received a sneak peek at this capability when the Auto-Trash team revealed its smart garbage can. It uses a camera to detect items placed inside its lid and sort them accordingly.

Improving conservation efforts

Instead of dispatching a fleet of human workers to analyze and count trees, for example, city officials could use drones and image recognition software to monitor the count, color and health of a natural area. WildTrack, a nonprofit, has already developed the Footprint Identification Technique to identify endangered species using only images of their footprints.

Using only images of footprints, a nonprofit called WildTrack has already developed the Footprint Identification Technique to identify endangered species. (Photo from Flickr.)

Once a system like WildTrack has collected a wealth of data, civic leaders can use the information to coordinate responses when plants or wildlife are threatened.

Despite common concerns of Big Brother watching our every movement, image recognition software is still far from ubiquitous. The technology’s myriad use cases make it impossible to ignore, and cities around the world are set to capitalize on its incredible value. For government officials and citizens alike, the spread of image recognition software is certain to be a godsend.

If you enjoyed this story, please recommend and share to help others find it! Feel free to leave a comment below.

The Mission publishes stories, videos, and podcasts that make smart people smarter. You can subscribe to get them here.