Training-free Zero-shot Composed Image Retrieval via Weighted Modality Fusion and Similarity
This addresses the challenge of reducing data collection and training costs for CIR, making it more accessible, though it is incremental as it builds on existing zero-shot methods.
The paper tackles the problem of composed image retrieval (CIR) by proposing a training-free zero-shot approach that avoids the need for supervised training or additional pretraining, achieving competitive performance on FashionIQ and CIRR datasets.
Composed image retrieval (CIR), which formulates the query as a combination of a reference image and modified text, has emerged as a new form of image search due to its enhanced ability to capture user intent. However, training a CIR model in a supervised manner typically requires labor-intensive collection of (reference image, text modifier, target image) triplets. While existing zero-shot CIR (ZS-CIR) methods eliminate the need for training on specific downstream datasets, they still require additional pretraining on large-scale image datasets. In this paper, we introduce a training-free approach for ZS-CIR. Our approach, Weighted Modality fusion and similarity for CIR (WeiMoCIR), operates under the assumption that image and text modalities can be effectively combined using a simple weighted average. This allows the query representation to be constructed directly from the reference image and text modifier. To further enhance retrieval performance, we employ multimodal large language models (MLLMs) to generate image captions for the database images and incorporate these textual captions into the similarity computation by combining them with image information using a weighted average. Our approach is simple, easy to implement, and its effectiveness is validated through experiments on the FashionIQ and CIRR datasets. Code is available at https://github.com/whats2000/WeiMoCIR.