TEMA: Anchor the Image, Follow the Text for Multi-Modification Composed Image Retrieval
For researchers in image retrieval, this work tackles the practical gap of handling multiple modifications in CIR, though the datasets are domain-specific and the architecture is incremental.
The authors address limitations in Composed Image Retrieval (CIR) with simple modification texts by introducing two multi-modification datasets (M-FashionIQ, M-CIRR) and proposing TEMA, a framework that handles both multi- and simple modifications. TEMA achieves superior retrieval accuracy and computational efficiency on four benchmarks.
Composed Image Retrieval (CIR) is an important image retrieval paradigm that enables users to retrieve a target image using a multimodal query that consists of a reference image and modification text. Although research on CIR has made significant progress, prevailing setups still rely simple modification texts that typically cover only a limited range of salient changes, which induces two limitations highly relevant to practical applications, namely Insufficient Entity Coverage and Clause-Entity Misalignment. In order to address these issues and bring CIR closer to real-world use, we construct two instruction-rich multi-modification datasets, M-FashionIQ and M-CIRR. In addition, we propose TEMA, the Text-oriented Entity Mapping Architecture, which is the first CIR framework designed for multi-modification while also accommodating simple modifications. Extensive experiments on four benchmark datasets demonstrate that TEMA's superiority in both original and multi-modification scenarios, while maintaining an optimal balance between retrieval accuracy and computational efficiency. Our codes and constructed multi-modification dataset (M-FashionIQ and M-CIRR) are available at https://github.com/lee-zixu/ACL26-TEMA/.