Predicting vision with deep learning
This initiative was awarded €10,000 in DigiShape seed money at the DigiShape Day on 15 October. Read the news item
PortVision is a proposal for a deep learning model to predict visibility in the port of Rotterdam, where fog and rain can (partially) shut down the port. The model combines satellite images, KNMI measurements and ECMWF forecasts with the Port Authority’s long-term visibility measurements and provides a forecast for the next 48 hours, including an estimate of reliability.
Forecasting fog is complex but essential, not only for the water sector but also for other sectors, such as the energy sector and the road sector. That is why we want to use PortVision to investigate how deep learning can link patterns in different datasets to improve visibility forecasts.
Purpose
Building a working prototype of an AI model that generates visibility forecasts by combining weather data, satellite images and local measurements for the Rotterdam port area.
Participating parties
HKV, Port of Rotterdam Authority, Pythia Energy Intelligence. New partners are welcome to join.
Call to the network
We invite everyone to join us and explore together how deep learning can help with better vision predictions and other complex parameters that are difficult to predict with traditional models.
In PortVision, we combine different datasets into one model to better understand rapidly changing local processes, and we are happy to brainstorm about this with partners.
Are you interested? Please contact us to join!
