Multi-Task Learning Using Uncertainty to Weigh Losses for Scene Geometry and Semantics
Numerous deep learning applications benefit from multi-task learning with multiple regression and classification objectives. In this paper we make the observation that the performance of...