qwenemb_cirr

该数据集包含多个配置，主要用于图像和文本检索任务。每个配置具有特定的特征、数据划分和大小。核心特征包括查询文本（'qry'）、查询图像路径（'qry_image_path'）、正例文本（'pos_text'）和正例图像路径（'pos_image_path'）。部分配置还包含环境标签（'env_label'）和实例标识符（'qry_inst', 'tgt_inst'）。数据集划分为原始集（'original'）或测试集（'test'），每个配置的样本数量和大小均有详细说明。例如，'cirr'配置包含28,225个样本，而'cirr_dev'配置包含4,181个测试样本。该数据集适用于多模态检索、图像文本匹配等任务。

This dataset comprises multiple configurations, primarily designed for image and text retrieval tasks. Each configuration features specific attributes, data splits and dataset sizes. The core attributes include query text ('qry'), query image path ('qry_image_path'), positive text ('pos_text') and positive image path ('pos_image_path'). Some configurations additionally provide environment labels ('env_label') and instance identifiers ('qry_inst', 'tgt_inst'). The dataset is split into the original set ('original') and test set ('test'), with detailed specifications for the sample count and size of each configuration. For example, the 'cirr' configuration contains 28,225 samples, while the 'cirr_dev' configuration includes 4,181 test samples. This dataset is suitable for multimodal retrieval, image-text matching and other related tasks.