DQE-CIR: Distinctive Query Embeddings through Learnable Attribute Weights and Target Relative Negative Sampling in Composed Image Retrieval

Researchers propose DQE-CIR, a new method for composed image retrieval that improves AI's ability to find images based on reference images and text modifications. The approach addresses limitations in current contrastive learning frameworks by using learnable attribute weights and target relative negative sampling to create more distinctive query embeddings.

• →DQE-CIR addresses semantic confusion and relevance suppression issues in existing composed image retrieval methods.
• →The method introduces learnable attribute weighting to better align language and vision features.
• →Target relative negative sampling selects more informative training examples by excluding easy negatives and false negatives.
• →The approach specifically improves performance on fine-grained attribute modifications in image retrieval.
• →This represents an advancement in multimodal AI systems that combine text and image understanding.