Exploring Instance Relations for Unsupervised Feature Embedding
This work addresses the limitation of neglecting instance relations in unsupervised feature embedding for computer vision tasks, offering an incremental improvement over existing contrastive learning methods.
The paper tackles the problem of unsupervised feature embedding by exploring instance relations, specifically intra-instance multi-view and inter-instance interpolation relations, to enhance contrastive learning methods. The proposed EIR approach achieves state-of-the-art or comparable performance on image classification and retrieval benchmarks.
Despite the great progress achieved in unsupervised feature embedding, existing contrastive learning methods typically pursue view-invariant representations through attracting positive sample pairs and repelling negative sample pairs in the embedding space, while neglecting to systematically explore instance relations. In this paper, we explore instance relations including intra-instance multi-view relation and inter-instance interpolation relation for unsupervised feature embedding. Specifically, we embed intra-instance multi-view relation by aligning the distribution of the distance between an instance's different augmented samples and negative samples. We explore inter-instance interpolation relation by transferring the ratio of information for image sample interpolation from pixel space to feature embedding space. The proposed approach, referred to as EIR, is simple-yet-effective and can be easily inserted into existing view-invariant contrastive learning based methods. Experiments conducted on public benchmarks for image classification and retrieval report state-of-the-art or comparable performance.