PromptStyler: Prompt-driven Style Generation for Source-free Domain Generalization

Open paper
Benchmark Model Rank Results
domain-generalization-on-domainnetPromptStyler (CLIP, ViT-L/14)#2Average Accuracy: 65.5
domain-generalization-on-domainnetPromptStyler (CLIP, ViT-B/16)#14Average Accuracy: 59.4
domain-generalization-on-domainnetPromptStyler (CLIP, ResNet-50)#19Average Accuracy: 49.5
domain-generalization-on-office-homePromptStyler (CLIP, ViT-L/14)#2Average Accuracy: 89.1
domain-generalization-on-office-homePromptStyler (CLIP, ViT-B/16)#9Average Accuracy: 83.6
domain-generalization-on-office-homePromptStyler (CLIP, ResNet-50)#17Average Accuracy: 73.6
domain-generalization-on-pacs-2PromptStyler (CLIP, ViT-L/14)#2Average Accuracy: 98.6
domain-generalization-on-pacs-2PromptStyler (CLIP, ViT-B/16)#7Average Accuracy: 97.2
domain-generalization-on-pacs-2PromptStyler (CLIP, ResNet-50)#14Average Accuracy: 93.2
domain-generalization-on-vlcsPromptStyler (CLIP, ViT-B/16)#6Average Accuracy: 82.9
domain-generalization-on-vlcsPromptStyler (CLIP, ViT-L/14)#8Average Accuracy: 82.4
domain-generalization-on-vlcsPromptStyler (CLIP, ResNet-50)#11Average Accuracy: 82.3

Rank counts only results with a code link.