JFT-300M Dataset

JFT-300M is an internal Google dataset used for training image classification models. Images are labeled using an algorithm that uses complex mixture of raw web signals, connections between web-pages and user feedback. This results in over one billion labels for the 300M images (a single image can have multiple labels). Of the billion image labels, approximately 375M are selected via an algorithm that aims to maximize label precision of selected images.

JFT-300M 是谷歌内部用于训练图像分类模型的数据集。图像标签采用一种算法进行标注，该算法融合了复杂的原始网络信号、网页间连接以及用户反馈。由此，对 3 亿张图像（单张图像可拥有多个标签）进行了超过十亿个标签的标注。在这十亿个图像标签中，大约 3.75 亿个标签是通过一种算法选出的，该算法旨在最大化所选图像标签的精确度。