Deep Semi-supervised Learning with Double-Contrast of Features and Semantics

TL;DR

This paper introduces an end-to-end deep semi-supervised learning method that enhances feature and semantic discrimination through double contrast, improving task-specific representations with theoretical support and benchmark validation.

Contribution

It proposes a novel double contrast approach for semantics and features in semi-supervised learning, addressing limitations of existing methods and providing theoretical explanations.

Findings

Effective in extracting discriminative features

Improves semantic consistency and discriminability

Validated on benchmark datasets

Abstract

In recent years, the field of intelligent transportation systems (ITS) has achieved remarkable success, which is mainly due to the large amount of available annotation data. However, obtaining these annotated data has to afford expensive costs in reality. Therefore, a more realistic strategy is to leverage semi-supervised learning (SSL) with a small amount of labeled data and a large amount of unlabeled data. Typically, semantic consistency regularization and the two-stage learning methods of decoupling feature extraction and classification have been proven effective. Nevertheless, representation learning only limited to semantic consistency regularization may not guarantee the separation or discriminability of representations of samples with different semantics; due to the inherent limitations of the two-stage learning methods, the extracted features may not match the specific downstream tasks. In order to deal with the above drawbacks, this paper proposes an end-to-end deep semi-supervised learning double contrast of semantic and feature, which extracts effective tasks specific discriminative features by contrasting the semantics/features of positive and negative augmented samples pairs. Moreover, we leverage information theory to explain the rationality of double contrast of semantics and features and slack mutual information to contrastive loss in a simpler way. Finally, the effectiveness of our method is verified in benchmark datasets.