Improved Training for Self-Training by Confidence Assessments

TL;DR

This paper proposes an improved self-training method that uses confidence assessments to leverage unlabeled data for training, applicable to classification and semantic segmentation tasks, reducing the need for extensive labeled datasets.

Contribution

It introduces a confidence-based approach for self-training that can be applied online to unlabeled data, enhancing learning efficiency in limited labeled data scenarios.

Findings

Effective on MNIST for classification

Successful semi-supervised segmentation on ADE20K

Improves training with minimal labeled data

Abstract

It is well known that for some tasks, labeled data sets may be hard to gather. Therefore, we wished to tackle here the problem of having insufficient training data. We examined learning methods from unlabeled data after an initial training on a limited labeled data set. The suggested approach can be used as an online learning method on the unlabeled test set. In the general classification task, whenever we predict a label with high enough confidence, we treat it as a true label and train the data accordingly. For the semantic segmentation task, a classic example for an expensive data labeling process, we do so pixel-wise. Our suggested approaches were applied on the MNIST data-set as a proof of concept for a vision classification task and on the ADE20K data-set in order to tackle the semi-supervised semantic segmentation problem.