Data Distillation: Towards Omni-Supervised Learning

Citations Over TimeTop 1% of 2018 papers

Abstract

We investigate omni-supervised learning, a special regime of semi-supervised learning in which the learner exploits all available labeled data plus internet-scale sources of unlabeled data. Omni-supervised learning is lower-bounded by performance on existing labeled datasets, offering the potential to surpass state-of-the-art fully supervised methods. To exploit the omni-supervised setting, we propose data distillation, a method that ensembles predictions from multiple transformations of unlabeled data, using a single model, to automatically generate new training annotations. We argue that visual recognition models have recently become accurate enough that it is now possible to apply classic ideas about self-training to challenging real-world data. Our experimental results show that in the cases of human keypoint detection and general object detection, state-of-the-art models trained with data distillation surpass the performance of using labeled data from the COCO dataset alone.

Related Papers

• → Semi-supervised learning: a brief review(2018)232 cited
• → Semi-Supervised Learning(2014)93 cited
• → Incorporate active learning to semi-supervised industrial fault classification(2019)35 cited
• → Recommender system designed using an ensemble approach to semi-supervised learning(2022)1 cited
• → Biodegradables(1994)