List your company

Vision-Language Alignment on Sugarcrepe swap-object

63.8AccuracyDCSM open-vocab
49.44853.17456.960.626Mar 10, 2025

Evaluation results

MethodLinks
DCSM open-vocabMode=Open-vocabulary2025.0363.8
CLIP ViT-B/16Backbone=ViT-B/162025.0360.2
DCSM synthTraining=Synthetic data2025.0359.8
CLIP ft.synthTraining=Fine-tuned on synthetic data2025.0350