We introduce GenSP, a data-driven framework that learns consistent spherical parameterizations across a collection of genus-0 shapes. Instead of optimizing the parameterization of each shape independently, our method learns a neural generative model that predicts a continuous map…
Adaptive video streaming is a widely used technique for delivering video content over the internet. One of the key challenges is determining the optimal encoding settings for each video, which can vary significantly based on its content and characteristics. In this paper, we prop…
Self-supervised learning (SSL) currently drives state-of-the-art performance in no-reference image quality assessment (NR-IQA). However, standard SSL pipelines uniformly apply synthetic distortions across the entire image field, which can limit their sensitivity to spatially loca…