Edition Sept 21st-23rd, 2018
DeepSnow is the first solution to combine AI and satellite images to generate data about ski resorts to minimize the effort to create the deep snow that we love.
"The goal of DeepSnow was to determine whether or not is possible to use Machine Learning / AI in combination with satellite images for the detection of infrastructure in ski resorts.
Creating and collecting relevant data, like “Where are the lifts? What’s the area of all ski slopes combined?”, etc. is a labor intensive and recurring task, binding human resources that are not available for the actual development of snow-making systems Techno Alpins Core business.
As time is limited in a Hackathon we focused and worked on the proof of concept and made it our goal to detect the outline of the ski slopes.
To achieve this goal we build a neural network for image segmentation, that takes a satellite image as an input and outputs the polygon shape of the ski slope in the picture. For usability we designed a React Frontend and a flask Backend. Challenges were the conceptualization of the task, determining what can be done (and what not!) in the course of one Hackathon.
After 2 nights with less than 4 hours of sleep DeepSnow is a running services that proves that the AI revolution does not stop in front of the mountains and that we will see intelligent systems in the operation/management/planning of ski resorts in the near future.
We believe the project has huge potential to become a mature application for TechnoAlpin and is not limited to its original use case as it can be expanded for example to help detect deforestation and flooding."