VisualAndTextualRepresentations

AbstractLearning visual and textual representations in the shared space from web-scale image-text pairs improves the performance of diverse vision-and-language tasks, as well as modality-specific tasks. Many attempts in this framework …...