KiloGram

该数据集是一个内容丰富的七巧板谜题集合，旨在研究抽象视觉推理。它不仅包含了分割图，还标注了各个部分的标签。数据集特别注重对74个七巧板进行密集标注，既包含了整体形状的标注，也包含了部分形状的标注。规模上，该数据集包含了1,016个七巧板，共13,404个注释，其任务是对抽象视觉推理进行评估。

This dataset is a comprehensive collection of tangram puzzles developed for researching abstract visual reasoning. It contains not only segmentation maps but also annotated labels for each constituent piece. The dataset places special focus on dense annotation for 74 tangram puzzles, with annotations covering both the overall shapes of the puzzles and the shapes of their individual components. In terms of scale, this dataset includes 1,016 tangram puzzles and a total of 13,404 annotations, and it is designed for the evaluation of abstract visual reasoning.

搜集汇总

背景与挑战

背景概述

KiloGram是一个七巧板（Tangram）多模态数据集，包含图像和描述数据，用于教育和研究目的。数据集提供JSON格式文件、预处理的训练数据以及微调模型，支持视觉和语言任务的结合分析。