Recent advances in semi-supervised learning (SSL) demonstrate that a combination of consistency regularization and pseudo-labeling can effectively improve image classification accuracy in the low-data regime. Compared to classification, semantic segmentation tasks require much more intensive labeling costs. Thus, these tasks greatly benefit from data-efficient training methods. However, structured outputs in segmentation render particular difficulties (e.g., designing pseudo-labeling and augmentation) to apply existing SSL strategies. To address this problem, we present a simple and novel re-design of pseudo-labeling to generate well-calibrated structured pseudo labels for training with unlabeled or weakly-labeled data. Our proposed pseudo-labeling strategy is network structure agnostic to apply in a one-stage consistency training framework. We demonstrate the effectiveness of the proposed pseudo-labeling strategy in both low-data and high-data regimes. Extensive experiments have validated that pseudo labels generated from wisely fusing diverse sources and strong data augmentation are crucial to consistency training for segmentation. The source code is available at https://github.com/googleinterns/wss.
翻译:半监督学习(SSL)的最近进展表明,一致性规范化和假标签的结合可以有效地提高低数据制度中图像分类的准确性。与分类相比,语义分割任务需要更加密集的标签费用。因此,这些任务从数据效率培训方法中受益匪浅。但是,结构化的分类产出特别难以(例如,设计假标签和增强)应用现有的SSL战略。为了解决这一问题,我们提出了一个简单和新颖的假标签设计,以生成结构清晰的假标签,用于用无标签或标签薄弱的数据进行培训。我们提议的伪标签战略是网络结构,不可在单阶段一致性培训框架内应用。我们证明拟议的伪标签战略在低数据和高数据系统中的有效性。广泛的实验证实,明智地使用多种来源和强力数据增强生成的假标签对于分类的一致性培训至关重要。源代码可在 https://github.com/googleinterns/ws查阅。