A U.S. utility used photographs taken by drones to identify equipment that needed repair across its far-flung distribution network. But manually examining these photos and opening a repair ticket was time-consuming and inefficient, and made it impossible to generate actionable intelligence in real time.
With a starting gallery of 1,000 labeled and classified drone images of equipment problems, this utility company needed properly trained analytics to build up, auto-classify and deliver actionable maintenance on their growing library of newly captured images. Because of Cognizant’s extensive background in image analytics, they engaged us for the project.
We used our BigDecision® AI platform to create an image analytics application, driven by artificial intelligence (AI), that assesses photos in real time and identifies problems such as broken or chipped insulators. Hosted on a cluster of high computing containers orchestrated by Cognizant partner Kubernetes, this self-service solution uses a real-time alerting engine to notify the utility’s maintenance team about needed repairs.
Image augmentation compensates for the lack of properly labeled images. It creates as many as 12 new labeled images from each original image by changing the lighting or angles or by adding new objects. This greatly increases the raw data from which the analytics application can learn, and thus its accuracy. We also automated critical activities such as data labeling, and building, training and deploying AI models.
The utility now has a fully managed data and analytics platform that enables data scientists to build, train and deploy AI models on-site or in the cloud. It greatly reduces the cost and time required for image analysis and needed repairs.
reduction in the effort required to scan images
problem identification and work order notifications to cut costs and speed up repairs
service levels, reduced service outages and improved customer experience
SPREAD THE WORD