Hive is a project and idea by Jaime Alemany from Lunático Astronomía.
Now that the CloudWatcher is able to measure so many meteorological data (ambient temperature, sky reflected IR, humidity, atmospheric pressure, rain, wind…) we are reactivating a project of old to apply machine learning to cloud and rain detection.
The plan is to collect a huge amount of data, CloudWatcher sensor data and all-sky images, preprocess it, and apply a self-learning algorithm.
With the invaluable help of our users, scattered all around the world, we plan:
- to be retrieving data for a whole year (at least!)
- to categorize the data, leaving a set of classified images (cloud cover, rain, etc) matched to a set of weather data.
- apply machine learning (to be determined, but most likely a CNN – convoluted neural network – and genetic algorithm) to be able to correctly identify clouds and rain from the sensor data.
We are going to need help at every step – we’ll devise a friendly front end, inspired on other citizen science project, to allow anyone to classify the information.