Mar 27, 2023 · We leverage contrastive learning by excavating hidden positives to learn rich semantic relationships and ensure semantic consistency in local regions.

Semantic segmentation is a major task for scene under- standing and plays a crucial role in many applications in- cluding medical imaging and autonomous driving ...

We leverage contrastive learning by excavating hidden positives to learn rich semantic relationships and ensure semantic consistency in local regions.

This work uses contrastive learning by excavating hidden positives to learn rich semantic relationships and ensure semantic consistency in local regions, ...

A gradient propagation to nearby similar patches, local hidden positives, is developed to learn local semantic consistency which is the nature of segmentation.

May 31, 2023 · Hyun Seok Seong, WonJun Moon, SuBeen Lee, Jae-Pil Heo, "Leveraging Hidden Positives for Unsupervised Semantic Segmentation", CVPR 2023.

... Leveraging hidden positives for unsupervised semantic segmentation enhances pixel grouping by identifying and utilizing implicit positive relationships ...

Mar 27, 2023 · Dramatic demand for manpower to label pixel-level annotations triggered the advent of unsupervised semantic segmentation.

Leveraging Hidden Positives for Unsupervised Semantic Segmentation, HP, CVPR2023 ; ACSeg: Adaptive Conceptualization for Unsupervised Semantic Segmentation ...

Specifically, we add the loss propagating to local hidden positives, semantically similar nearby patches, in proportion to the predefined similarity scores.