目标检测

DQEN proposes a dual query enhancement network for DETR-based human-object interaction detection. The method enhances object queries and interaction queries to improve the discrimination and understanding of human-object interaction representations.

In this paper, a novel pseudo-margin evaluation algorithm is designed to address the quality problem of pseudo-labels brought by those new novel instances.

This brief proposes a new algorithm using attributes based category similarity for accurate alignment between visual features and semantic concepts.