Applying sensing and AI to innovative problem solving
Sprawling municipalities face many challenges. Traffic congestion slows productivity, contributes to air quality problems, and wastes energy. Crowds of people bring concerns over public health, safety, and security. Distributed sensors help monitor conditions, but in turn generate volumes of big data. AI inference helps cut through routine monitoring, checking compliance, spotting problems, or highlighting opportunities for savings.
Urban environments can be a great setting for people watching, but in some cases a more quantitative approach is needed. Smart surveillance can apply to any gathering place in a city - rooms in a conference center, streets and walkways, and special events.
Networks of cameras provide high-resolution video from many angles, and volumes of data. The task of watching all that data would numb human operators, opening the chance that something important is missed. AI inference is proving valuable in accurate recognition of people and items they wear or carry.
Simple AI inference models are now highly accurate at smart surveillance. Tasks like verifying room capacity, checking mask compliance, and assessing foot traffic can be automated, allowing human resources to be deployed for other priorities.
Traffic is a concern as cities grow and freeways and thoroughfares become more congested. A side effect of increased traffic is difficulty finding parking. Helping traffic flow more efficiently is a good application for AI inference in spotting patterns.
Traffic congestion also impacts the perceived livability of a city. It increases commute times, drives up fuel consumption, adds to air pollution, and frustrates drivers. Building more roads or adding lanes may alleviate problems in one area, but the point of congestion usually moves to another spot. More timely information is needed.
Traffic management systems gather information from many sources - like cameras, in-road sensors, vehicle-to-everything (V2X) systems - creating a better picture of what is happening in real-time. With that data, signals can be controlled, signage can indicate alternate routes, tolls can be automated, keeping vehicles moving.
Retailers are always looking for ways to enhance the experience of their shoppers, and increase revenue per visit. Smart retail systems with AI inference help match shoppers to goods and entice them with offers tailored for their preferences.
Smart retail systems can help guide shoppers from the moment they enter the store to items of interest. Along the way, shoppers might pass digital signage displays with cameras gathering more non-personally identifiable information - gender, age range, clothing and accessories. With that information, dynamic offers can be presented.
Smart retail may also work with other sensors, like RFID-enabled shopping carts that sense what items a shopper has selected, making complementary suggestions. Traffic flow and signage watch time are also important metrics for optimizing placements.
AI inference helps faster decision-making from data
- MERA Software Framework for AI takes models from standard machine learning platforms including PyTorch, TensorFlow, and ONNX.
- SAKURA ASIC delivers 40 TOPS in 10W of power, and up to five chips can be tied together via PCIe in a configuration for 200 TOPS.
- The same DNA IP in the SAKURA ASIC can be configured into fast FPGA AI accelerator cards for servers, adding reconfigurability to high performance.