Attribute-Specific Fashion Retrieval (ASFR) focuses on retrieving images based on fine-grained, attribute-specific criteria rather than naive global visual similarity, enabling more precise and interpretable search results. Existing ASFR methods ideally assume that all attribute semantics are in-domain distributions of the training datasets. However, realistic scenarios are generally more complex and naturally contain unseen attribute information, often resulting in ungeneralizable retrieval outcomes. In this paper, we take the first step to address the new and challenging open-world ASFR setting, which involves handling diverse and practical attributes instead of relying solely on predefined attribute sets in closed-world scenarios. Specifically, to comprehend unseen attributes, we propose a novel LLM-based Commonsense Knowledge Infusion (CoKi) framework that integrates commonsense knowledge as complementary context into attribute representations using a Large Language Model (LLM). By infusing such LLM-based commonsense knowledge through descriptive contexts, our method enables robust semantic enrichment and effective generalization to unseen attributes. Additionally, we introduce a modality-switchable prompt and an imputation mechanism to ensure model robustness across diverse input configurations by dynamically adapting to missing modalities. Extensive experiments demonstrate that our approach not only achieves state-of-the-art in-domain retrieval performance but also significantly enhances adaptability to unseen attributes and cross-domain generalization, establishing a new benchmark for fine-grained fashion retrieval in open-world scenarios. Our source code is publicly available at https://github.com/HuiGuanLab/CoKi.

Vector image is a type of image composed of many geometric primitives. Compared with bitmaps, vector images have the ability to save memory as well as to enlarge without distortion. Meanwhile, it has been commonly adopted in data visualization (image data) because it can be scaled to multiple sizes to fit different scenes. For instance, it can be applied for the illustrations in newspapers and magazines, the logo on the web, the background for poster, the design of text, and traffic signs. However, transforming a bitmap to vector image is still a challenging problem because of the complicated content of a bitmap, which tends to consist of more than just simple geometry. Aiming at this issue, there is a new approach proposed to transform from bitmaps to vector images, which is based on triangle units and consists of three steps. In detail, firstly, there is an initial mesh constructed for one image in pixel level after detecting features. Then, the initial mesh will be simplified by collapsing two vertices as the initial mesh is too dense to represent one image. Specifically, there are two main parts in this step, which are collapse conditions and collapse influences. In the step of collapsing, issues such as overlap and sharp triangles can be conquered by a sort-edge method (which will be illustrated specifically later). The final step is to select one color for each triangle, since it is helpful to save the memory and speed up the process of this method. In addition, one color will represent one triangle; hence, in the final step, the four-triangle sample method will be applied in order to prevent a vector image from generating too large color discontinuity. Once the pretest proceeds without mistake, the method above is able to work for the general bitmaps. Note that our method can be applied to information security and privacy, since one image can be encoded to some triangles and colors.