Gauge and Dial Reading with Computer Vision - lilz

Remote Monitoring with LiLz: A Machine Learning Solution

The task of physically inspecting gauges and valves in remote or difficult-to-access locations is one that no maintenance professional relishes. LiLz offers a solution, enabling remote oversight of these physical interfaces through a sophisticated application of machine learning.

This Japanese startup, originating from Okinawa, has been developing its technology for some time – previously covered by TC JP. Despite the clear benefits of its service, widespread adoption has been gradual. LiLz showcased its innovation at CES as part of a larger delegation representing the country’s trade organizations.

Device Overview

The LiLz device resembles a compact tablet, devoid of a display screen. It integrates a camera, lighting, processing chips, and communication modules, powered by a substantial battery capable of lasting up to three years.

The device is mounted to provide a clear view of the target gauge or dial. Following initial picture and signal confirmation, the device is configured via an app to interpret the visual data. It can accurately read circular, semicircular, and linear gauges, as well as digital, rolling, and analog displays, and even interpret colored warning indicators. (The underlying ML is complex – this article provides further insight.)

Upon setup, the device transmits readings in real-time or at pre-defined intervals to a centralized dashboard. Data accessibility is also provided through an API, allowing for querying and recording elsewhere. Data transmission occurs via LTE or Bluetooth connectivity.

Image Credits: LiLz

Target Applications

This solution is specifically designed for infrastructure and heavy industrial settings, where legacy equipment is often situated in challenging environments. These include rooftops, underground locations (with sufficient signal penetration), and complex factory or warehouse layouts.

Manual inspection of these dials is not only a monotonous task but can also present safety hazards. While robotic automation is an alternative, a network of IoT devices appears to be a more pragmatic approach than deploying a constantly moving quadrupedal robot.

Company Growth and Expansion

LiLz CTO, Kuba Kolodziejczyk, reports significant growth since the company’s launch in 2020 and a $2.2 million Series A funding round in early 2021.

Image Credits: LiLz

“From 240 cameras deployed across 34 locations with a limited number of initial users prior to June 2020, we have expanded to 2000 cameras operating at 320 locations for over 100 clients. We anticipate reaching 5000 cameras by the end of this year,” he stated in correspondence with TechCrunch. “Currently, several clients utilize more than one hundred cameras at a single location.”

Initially focused on building management, LiLz has broadened its reach to include chemical and industrial plants, construction sites, manufacturing facilities, and public infrastructure, responding to evolving customer needs.

Ongoing Development

The device’s capabilities are continually being enhanced, primarily through software updates. These improvements focus on remote update functionality, increased accuracy, resilience to interference, and the addition of data sharing features. A new, explosion-proof hardware version is also now available.

Future product development includes a “sound search” feature and an object counting function within the camera’s field of view, though these are currently in early stages. Additional monitoring capabilities, such as float and level gauges, are also planned.

LiLz is currently leveraging CES to connect with potential clients outside of Japan. While the event’s overall impact remains to be seen, it may spark interest among technicians and managers seeking a convenient and effective monitoring solution. Further information can be found on their recently launched English-language website.