
提案されたエンコーダー専用アーキテクチャは、オープンボキャブラリーの視覚関係検出において強力なパフォーマンスを達成し、従来のデコーダーに基づくアーキテクチャよりも優れていることを示しています。


coremsg

scene-graph-vit-efficient-visual-relationship-detection-architecture


Scene-Graph ViT: Efficient Visual Relationship Detection Architecture


title_rewrite


Proposing Groupwise Query Specialization and Quality-Aware Multi-Assignment (SpeaQ) to enhance Transformer-based Visual Relationship Detection models by training specialized queries and providing richer training signals.


groupwise-query-specialization-and-quality-aware-multi-assignment-for-transformer-based-visual-relationship-detection


Groupwise Query Specialization and Quality-Aware Multi-Assignment for Transformer-based Visual Relationship Detection
