“Self-supervised learning for inverse problems”
Politecnico di Torino, Italy
Self-supervised learning allows to use the data themselves to extract supervisory signals for training, thus reducing data requirements and circumventing the need for extensive ground truth labels. This is particularly interesting in the solution of inverse problems in imaging (e.g., denoising, super-resolution, …) where ground truth data can be hard or impossible to collect. The candidate will work on improving current self-supervised techniques by addressing their current limitations. Publications at top machine learning conferences (NeurIPS, ICLR, CVPR, …) as well as prestigious image processing journals (IEEE TIP, IEEE TGRS, …) are expected. The ideal candidate has a PhD with several publications on machine learning topics.
The candidate will join the Image Processing and Learning (IPL) Lab at Politecnico di Torino. IPL has expertise on inverse problems including denoising, super-resolution, compressed sensing, etc. in the fields of computer vision and remote sensing and publishes at top machine learning and signal processing venues. Some IPL members are also members of the ELLIS Society, fostering networking in the European machine learning scene. The contract is in the context of a European project and has a duration of 1 year (extendable). Contract type is “assegno di ricerca”, with salary to be negotiated depending on expertise. In view of the current COVID-19 situation, remote onboarding and working conditions will be considered until the situation returns to normal.
For more information, contact Prof. Enrico Magli (enrico.magli AT polito.it).