LEARNING TO DECOMPOSE VISUAL FEATURES WITH LATENT TEXTUAL PROMPTS

Keyphrases

• Visual Features 100%
• Vision-language 100%
• Zero-shot 100%
• Vision-language Models 100%
• Textual Prompt 100%
• Recent Advances 50%
• Visual Representation 50%
• Encoder 50%
• ImageNet 50%
• Dual Model 50%
• Language Input 50%
• Large Margin 50%
• Linear Layer 50%
• Model Driven Architecture 50%
• Test Accuracy 50%
• Linear Probe 50%
• ResNet50 50%
• Class Name 50%
• Prompt Tuning 50%
• Vision-language Alignment 50%

Computer Science

• Visual Feature 100%
• Language Modeling 100%
• Visual Representation 50%
• Language Input 50%
• Model Architecture 50%
• Residual Neural Network 50%
• Prompt Tuning 50%