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

Benchmark Model Rank Results
scene-graph-generation-on-visual-genomeSpeaQ (without reweighting)#1Recall@50: 32.9Recall@100: 36.0mean Recall @100: 14.1R@100: 36.0…
scene-graph-generation-on-visual-genomeSpeaQ (with reweighting)#2Recall@50: 32.1Recall@100: 35.5mean Recall @100: 17.6R@100: 35.5…