Forget cats, this neural network spots solar panels

        December 19, 2018 IEEE Spectrum | United States | renewables: solar

A Stanford University machine-learning algorithm has analyzed satellite images to count the number of solar panels across the US. Correctly identifying 93% of panels in 370,000 100-square-foot images, DeepSolar counted almost 1.5 million solar installations. Tied to GPS and census data, future uptake can be predicted.