Multi-Task Learning Using Uncertainty to Weigh Losses for Scene Geometry and Semantics
This addresses the challenge of manual weight tuning in multi-task learning for computer vision applications, offering a practical improvement.
The paper tackles the problem of tuning loss weights in multi-task learning by proposing a method that uses task uncertainty to automatically weigh losses, enabling joint learning of depth regression and segmentation from monocular images, and it outperforms single-task models.
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 such systems is strongly dependent on the relative weighting between each task's loss. Tuning these weights by hand is a difficult and expensive process, making multi-task learning prohibitive in practice. We propose a principled approach to multi-task deep learning which weighs multiple loss functions by considering the homoscedastic uncertainty of each task. This allows us to simultaneously learn various quantities with different units or scales in both classification and regression settings. We demonstrate our model learning per-pixel depth regression, semantic and instance segmentation from a monocular input image. Perhaps surprisingly, we show our model can learn multi-task weightings and outperform separate models trained individually on each task.