ICFG-PEDES

该数据集包含图像及其相应的文本描述，重点在于个体身份和细粒度信息的强调。数据集被划分为训练集（包含34,674个图像-文本对）和测试集（包含19,848个图像-文本对）。描述文本的平均长度为37个单词，词汇量包含5,554个不同的单词。该数据集总共包含54,522张图像及其对应的文本描述，覆盖了4,102个不同的身份。其任务是进行文本到图像的人物检索。

This dataset contains images and their corresponding textual descriptions, with an emphasis on individual identities and fine-grained information. The dataset is divided into a training set (34,674 image-text pairs) and a test set (19,848 image-text pairs). The average length of the descriptive texts is 37 words, and the vocabulary consists of 5,554 distinct words. In total, the dataset includes 54,522 images and their corresponding textual descriptions, covering 4,102 unique identities. The targeted task for this dataset is text-to-image person retrieval.

搜集汇总

背景与挑战

背景概述

ICFG-PEDES是一个用于文本到图像部分感知行人重识别的数据集，收集自MSMT17并保持其原始存储结构。数据集的具体划分由ICFG-PDES.json文件决定，适用于训练和测试相关模型。