Machine Learning
Case Studies

Our client needed to identify and classify power transmission poles.  Their process was to fly the line of towers and then manually select an image for each pole for the file.

The InfraSix Solution was to use the images they had captured to execute a machine learning program to automatically identify and classify the transmission poles, crop the photos to the proper proportions and geo-tag each pole.  We were then able to use hundreds of miles of images and complete a report in just a matter of hours with high accuracy.  

Our client needed to know if commercial roofs had water under the roof membrane.  Their process was to walk the roof looking for visual cues, or find leaks in the ceiling indoors.

The InfraSix Solution was to use a drone equipped with a highly accurate IR Thermal Camera/HD cameral to fly the roof at sunset.  The thermal sensor was able to detect thermal anomalies indicating likely water penetration. The roofing client then drilled core samples to verify the presence of moisture and allow us to calibrate for actual moisture vs. false positives.

Our client needed to have record of the size, slope and shape of the roofs on commercial buildings.  Their process was to either walk the roofs or fly a drone to capture images, which was an expensive proposition over thousands of roofs.  

The InfraSix Solution was to use satellite imagery as the image source for Machine Learning to identify, classify and measure commercial roofs.  We created Machine Learning software to analyze up to 1,000 roofs at a time, use geo-tags to identify specific buildings of interest, and gain the ability to triage which roofs may warrant a closer inspection—making the entire process far more efficient.