Computer Vision and Machine Learning
Field-Object Classifier
A Global Power Research Institute
In Partnership with Black Ice, LLC
  • Computer Vision
  • Machine Learning
  • Field analysis acceleration
The Problem
On transmission power lines, much of the equipment was installed long ago and a service engineer can have a challenge determining the make and model of an asset and what specific issues to be aware of for that particular asset. Our research-institution client wanted to develop a proof-of-concept app to explore accelerating the process of identifying and inspecting certain types of electric-power-infrastructure field objects.
The Solution
YoJonesy developed a mobile application that uses computer vision (“CV”) and machine learning (“ML”) to automatically identify objects within a certain class and present the user with information about it, including make, model, and specific types of faults that the particular model may be prone to develop.
Strategy
Since we knew we’d be attempting to maximize ML image classification accuracy among objects that looked quite similar, we decided to take a 2-pronged approach. Rather than attempt cross-platform capabilities from the start (as we usually do), we opted to fully explore the latest iOS ML capabilities on one path, and to fully explore the Google Tensorflow/Keras capabilities on a second path.

This approach allowed us to maximize the performance of each platform. We used Apple’s CoreML, for example, with supplemental training and tweaking of our custom model, and this resulted in an application that can accurately discern between 3 very similar objects with few notable differences.
Key Takeaways
  • Sometimes, it’s necessary to take the specialized approach, especially when on the cutting edge of a new technology
  • AI and ML are advancing so rapidly that we absorbed multiple generations of change during this one relatively short project
Get in touch
Have ideas or questions? Let’s talk. We’d love to help.
Contact YoJonesy