Bootstrapping Semantic Segmentation with Regional Contrast
This work addresses the challenge of data efficiency in semantic segmentation, particularly for applications with limited labeled data, though it is incremental as it builds on existing segmentation networks.
The paper tackles the problem of semantic segmentation by introducing ReCo, a regional contrastive learning framework that improves performance in both semi-supervised and supervised settings, achieving high-quality models with only 5 examples per semantic class.
We present ReCo, a contrastive learning framework designed at a regional level to assist learning in semantic segmentation. ReCo performs semi-supervised or supervised pixel-level contrastive learning on a sparse set of hard negative pixels, with minimal additional memory footprint. ReCo is easy to implement, being built on top of off-the-shelf segmentation networks, and consistently improves performance in both semi-supervised and supervised semantic segmentation methods, achieving smoother segmentation boundaries and faster convergence. The strongest effect is in semi-supervised learning with very few labels. With ReCo, we achieve high-quality semantic segmentation models, requiring only 5 examples of each semantic class. Code is available at https://github.com/lorenmt/reco.