Solar Panel Detection
Reply has developed a solution for a company in the financial industry to detect solar panels on house roofs.
#Convolutional Neuronal Networks
# Insurance Industry
# Class Activation Map
# Visualization
,allowExpansion)
Neural Networks for new Business and Customer Loyalty in Financial Services
Reply has developed a solution for a company in the financial industry to detect solar panels on house roofs. The solution uses satellite images for this purpose, which are evaluated with the help of image recognition processes and neural networks. The aim of the project was to identify existing customers with insufficiently insured solar installations, and to inform them of the best possible insurance to protect the valuable installations.
The technical implementation by Reply was carried out with a deep learning model that recognises solar plants on aerial photographs. For this purpose, the addresses of those with insurance were converted into coordinates in a first step. Then, an aerial image was requested from a service provider in a small radius around this coordinate. Since this process is fully automated, a high degree of accuracy is required. It must be ensured that, for example, a solar system installed on the neighbouring roof does not lead to an incorrect result.
Deep learning and Artificial Intelligence for unambiguous results
To train a Deep Learning model, a large amount of training data is first required. For this purpose, several thousand images of houses were sorted into images with and without solar panels. On this basis, the model was able to learn to distinguish solar installations from other structures. Seen from the air, for example, conservatories and the wooden beams of a carport look very similar to a solar installation.
Visualisation methods from the field of artificial intelligence are helpful in recognising such difficult structures. If one follows the path of an image through the model, it becomes clear through which image areas the model arrived at its decision. This makes it possible to identify errors and correct them in a targeted manner by expanding the training data.
Strengthening customer loyalty
Reply was able to provide the insurance company with a model that shows, for any address in Germany, whether the house has a solar system on the roof. The model was applied to all of the insurance company's existing customers to proactively inform them of incomplete insurance coverage, thus strengthening customer loyalty.
The project was presented by Deutsche Bank in a video marking the occasion of the 150th anniversary of Deutsche Bank. The series #Expeditiown150 is dedicated to new trends and innovations from the financial industry.
)
Machine Learning Reply is the Reply Group company that specialises in Machine Learning, Cognitive Computing and Artificial Intelligence solutions based on the Google technology stack. Based on the latest developments in artificial intelligence, Machine Learning Reply applies innovative Deep Learning, Natural Language Processing and Image/Video Recognition techniques to different realms of application, such as smart automation, predictive engines, recommendation systems and conversational agents.