Cross-modal Learning by Hallucinating Missing Modalities in RGB-D Vision
Nuno C. Garcia⁎,†; Pietro Morerio⁎; Vittorio Murino⁎,‡ ⁎Pattern Analysis & Computer Vision (PAVIS), Istituto Italiano di Tecnologia (IIT), Genova, Italy†Universita' degli Studi di Genova, Genova, Italy‡Universita' degli Studi di Verona, Verona, Italy
Abstract
Diverse input data modalities can provide complementary cues for several tasks, usually leading to more robust algorithms and better performance. However, while a (training) dataset could be accurately designed to include a variety of sensory inputs, it is often the case that not all modalities are available in real life (testing) scenarios, when the model is to be deployed. This raises the challenge of how ...
Get Multimodal Scene Understanding now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
