As a PhD student, you will investigate the potential of human-machine interaction in remote sensing. In particular, you will study the potential of systems based on visual question answering (VQA) for information retrieval from satellite images. The system researched would allow anyone (from laymen to experts) to query remote sensing images in search of relevant information, by asking questions in English about spatio / temporal varying topics of interest. Possibility of including data from multiple sources (historical maps, social media) will also be at the core of the research.
The system will be tested in a series of monitoring tasks in Alpine environments, in particular towards aspects related to quantification and detection of changes in natural environments, tourism attractivity and/or yield prediction. It will lead to a demonstrator
Main duties and responsibilities:
- Perform data acquisition of large amounts of satellite data over the Swiss Alps.
- Create a database of spatio/temporal image-question-answer triplets.
- Develop machine learning-based methodologies to analyze the data and create a demonstrator of the potential of spatio/temporal remote sensing VQA.
- Test alternative solutions, in particular to improve the language model.
- Write publications.
- Attend international conferences.
- Participate to education of the ENAC