Continuous Object Representation
Max discusses the shift from discrete pixel grids to continuous object representation in computer vision models, highlighting the importance of sampling an underlying continuous signal. The use of gaussian processes allows for modeling data with uncertainty, enabling subpixel sampling and working with sparse locations in a novel way.In this clip
From this podcast
Machine Learning Street Talk (MLST)
#036 - Max Welling: Quantum, Manifolds & Symmetries in ML
Related Questions