Weather webcams continuously take pictures of the sky and landscape for meteorologists and the general public to get an impression of the current weather situation. They are a great tool to verify the forecast and see the local deviation.
For this project we would like to classify the types of clouds that are visible in the images and what the weather situation currently is. Is it sunny? Are we seeing rain clouds?
You will be using machine learning (eg. auto-encoders) and dimensionality reduction techniques (eg. t-SNE, PCA) to find clusters in the images. These groupings mean that similar clouds / weather conditions are depicted in the images. You will look at self-supervised techniques in order to minimize the amount of manual labelling necessary.
We have a large collection (16+ million) of webcam images from the Czech Meteorological Service (CHMI) that covers 98 locations over 18+ months in 5 minute intervals. This dataset can be a valuable asset to the research community, if there is proper annotation and meta-data for each image available. Your thesis will contribute to this list of additional knowledge we have over the images and help researchers to train better models with this data in the future.