Comprehensive Fashion Understanding from Images through Multimodal LLMs
Journal of Design Intelligence · 24.04.2026 · 1 (2) · 1-15
DOI: 10.64509/jdi.12.54 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Design Intelligence, online erschienen am 24.04.2026. Verfasst von Yanghong Zhou und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
With the popularity of social media platforms in the information era, people are sharing a large volume of digital content online, including photos and other media data. The availability of media data over the Internet makes it very attractive to mine useful information from these data. However, for fashion image recognition, most existing datasets and methods are limited to coarse-grained categories or a small subset of attributes, lacking a comprehensive dataset and a holistic understanding of clothing. In this paper, we construct a fashion dataset with structured and comprehensive clothing descriptions using an MLLM (Multimodal Large Language Model) grounded in fashion knowledge. To ensure accurate understanding by the MLLM, we define a fashion schema and propose a schema-guided prompting strategy. Leveraging this dataset, we train a model based on BLIP (Bootstrapping Language-Image Pre-training) to recognize category information and fine-grained attributes of clothing images, and to learn text–image aligned representations for clothing image retrieval. Experimental results demonstrate that our approach significantly improves both attribute recognition and cross-modal retrieval performance compared to existing baselines.
Zitieren
APA
Zhou, Y., Tian, H., Chen, Y., & Mok, P. Y. (2026). Comprehensive Fashion Understanding from Images through Multimodal LLMs. Journal of Design Intelligence, 1(2), 1-15. https://doi.org/10.64509/jdi.12.54
Vancouver
Zhou Y, Tian H, Chen Y, Mok PY. Comprehensive Fashion Understanding from Images through Multimodal LLMs. Journal of Design Intelligence. 2026;1(2):1-15. doi:10.64509/jdi.12.54
BibTeX
@article{Zhou2026,
author = {Zhou, Yanghong and Tian, Hao and Chen, Yang and Mok, P. Y.},
title = {Comprehensive Fashion Understanding from Images through Multimodal LLMs},
journal = {Journal of Design Intelligence},
year = {2026},
volume = {1},
number = {2},
pages = {1--15},
doi = {10.64509/jdi.12.54}
}
EUVIMED