P9D

该数据集名为P9D，是一个全面且统一的产品图像-文本基准数据集，包含了来自9个行业的超过一百万个产品图像-文本对。它旨在支持视觉-语言连续预训练（VLCP）的研究。除了包含1,014,599对图像-文本用于训练，以及2,846对用于测试外，该数据集还涵盖了超过3,800个类别，分为9个任务，以模拟现实世界的长尾分布。其规模超过了一百万图像-文本对，任务重点在于视觉-语言连续预训练。

The dataset named P9D is a comprehensive and unified product image-text benchmark dataset containing over one million image-text pairs from 9 industries. It is designed to support research on Vision-Language Continuous Pre-training (VLCP). In addition to 1,014,599 image-text pairs for training and 2,846 pairs for testing, the dataset covers more than 3,800 categories divided into 9 tasks to simulate real-world long-tailed distributions. Boasting a scale of over one million image-text pairs, the dataset focuses on Vision-Language Continuous Pre-training research.