Fabric-Image-Data (FID)
收藏资源简介:
Fabric-Image-Data (FID) 数据集包含了12,181张羊毛面料图像,分为四种类型:格子、图案、纯色和条纹。该数据集旨在用于织物图像的分类,但本文将其用于文本到图像的跨模态检索任务。为了支持这一任务,作者为每张图像生成了两种类型的文本描述:自由形式的自然语言描述和基于属性的模板描述。这些描述是通过使用大型语言模型自动生成的,以评估不同视觉语言模型在检索性能上的表现。数据集的创建过程包括使用ChatGPT-4o-mini生成描述,并使用CLIP、LAIONCLIP和Perception Encoder三种视觉语言模型进行检索性能评估。该数据集的应用领域是织物行业,旨在解决基于文本描述检索织物图像的问题。
The Fabric-Image-Data (FID) dataset contains 12,181 wool fabric images categorized into four types: plaid, patterned, solid-color, and striped. While originally intended for fabric image classification, this dataset is employed for text-to-image cross-modal retrieval in this paper. To support this task, the authors generated two types of text descriptions for each image: free-form natural language descriptions and attribute-based template descriptions. These descriptions were automatically generated using large language models to evaluate the retrieval performance of different vision-language models. The dataset construction process involved generating descriptions via ChatGPT-4o-mini, and conducting retrieval performance evaluation using three vision-language models: CLIP, LAIONCLIP, and Perception Encoder. Targeted for the textile industry, this dataset aims to address the challenge of retrieving fabric images based on text descriptions.
Fabric Image Data (FID) 数据集概述
数据集基本信息
- 名称: Fabric Image Data (FID)
- 用途: 纺织品图像分类
- 数据量: 12181张羊毛织物图像
- 图像尺寸: 420×570像素
- 下载链接: https://pan.baidu.com/s/1qeUYiGd4FQ3PcByu67do_w
- 提取密码: oj4y
数据分类
- lattice: 3128张图片
- pattern: 768张图片
- solid: 4169张图片
- stripe: 4116张图片
引用信息
如果FID数据集对您的研究有帮助,请考虑引用以下文献: Liu, R., Yu, Z., Fan, Q. et al. The improved method in fabric image classification using convolutional neural network. Multimed Tools Appl (2023). https://doi.org/10.1007/s11042-023-15573-w
示例图像
- lattice:

- printed:

- solid:

- stripe:





