MIP-GAF
收藏资源简介:
MIP-GAF数据集由印度理工学院鲁尔基分校创建,旨在识别图像中最重要的人物(MIP)。该数据集包含16550张图像,涵盖多种社交场景,如家庭聚会、节日庆祝、街头斗殴等。数据集通过多模态大语言模型(MLLM)进行标注,确保了标注的准确性和上下文理解。创建过程包括使用MLLM进行初步标注,然后由人工进行验证和分类。MIP-GAF数据集主要用于图像字幕生成、社交关系分析等领域,旨在解决在复杂社交场景中识别最重要人物的挑战。
The MIP-GAF dataset was developed by the Indian Institute of Technology Roorkee, with the primary objective of identifying the most significant individual (MIP) within images. This dataset consists of 16,550 images spanning diverse social scenarios, including family gatherings, holiday celebrations, street altercations, and others. The dataset was annotated using multimodal large language models (MLLMs) to guarantee annotation accuracy and contextual understanding. Its construction workflow includes preliminary annotation via MLLMs, followed by manual verification and categorization. The MIP-GAF dataset is mainly utilized in domains such as image caption generation and social relationship analysis, aiming to tackle the challenge of identifying the most important person in complex social scenes.




