Zobrazeno 1 - 2
of 2
pro vyhledávání: '"Tater, Tarun"'
The visual representation of a concept varies significantly depending on its meaning and the context where it occurs; this poses multiple challenges both for vision and multimodal models. Our study focuses on concreteness, a well-researched lexical-s
Externí odkaz:
http://arxiv.org/abs/2410.11657
Existing deep learning approaches for learning visual features tend to overlearn and extract more information than what is required for the task at hand. From a privacy preservation perspective, the input visual information is not protected from the
Externí odkaz:
http://arxiv.org/abs/2005.10220