Textual-Visual-Semantic-Dataset

Textual-Visual-Semantic-Dataset是一个基于COCO Captions扩展的数据集，旨在通过添加场景信息（如图像中的物体）来增强图像描述的文本视觉上下文。该数据集包含413,915条训练数据和87,721条验证数据，可用于提升图像描述系统中的文本相似性和语义相关性。数据集的创建过程涉及使用现成的工具从图像中提取视觉概念，并结合文本信息进行标注。该数据集适用于计算机视觉和自然语言处理领域，特别是在需要结合视觉和语言理解的任务中，如图像描述生成和文本相似度计算。

Textual-Visual-Semantic-Dataset is an extended dataset based on COCO Captions, aiming to enhance the textual-visual context of image captions by adding scene information such as objects in images. This dataset contains 413,915 training samples and 87,721 validation samples, which can be used to improve textual similarity and semantic relevance in image captioning systems. The dataset construction process involves using off-the-shelf tools to extract visual concepts from images and combine them with textual information for annotation. This dataset is applicable to the fields of computer vision and natural language processing, especially for tasks requiring integrated visual and language understanding, such as image caption generation and text similarity computation.