Many prediction tasks contain uncertainty. In some cases, uncertainty is inherent in the task itself. In next-frame or future prediction, for example, many distinct outcomes are equally valid. In other cases, uncertainty arises from the way data is labeled. For example, in object detection, many objects of interest often go unlabeled, and in human pose estimation, occluded joints are often labeled with ambiguous values. In this work we focus on a principled approach for handling such scenarios. In particular, we propose a framework for reformulating existing single-prediction models as multiple hypothesis prediction (MHP) models, and we propose an associated meta loss and optimization procedure to train them. To demonstrate our approach, we consider three diverse applications: human pose estimation, future prediction and image classification. We find that MHP models outperform their single-hypothesis counterparts in all cases, and that MHP models simultaneously expose valuable insights into the variability of predictions.     «

Many prediction tasks contain uncertainty. In some cases, uncertainty is inherent in the task itself. In next-frame or future prediction, for example, many distinct outcomes are equally valid. In other cases, uncertainty arises from the way data is labeled. For example, in object detection, many objects of interest often go unlabeled, and in human pose estimation, occluded joints are often labeled with ambiguous values. In this work we focus on a principled approach for handling such scenarios....     »