HICO-DET

HICO-DET 是用于检测图像中人与物体交互 (HOI) 的数据集。它包含 47,776 张图像（训练集中 38,118 幅，测试集中 9,658 幅），由 80 个对象类别和 117 个动词类别构成的 600 个 HOI 类别。 HICO-DET 提供超过 15 万个带注释的人-物对。 V-COCO 提供 10,346 张图像（2,533 张用于训练，2,867 张用于验证，4,946 张用于测试）和 16,199 个人物实例。每个人都有 29 个动作类别的注释，并且没有包括对象在内的交互标签。

HICO-DET is a dataset for human-object interaction (HOI) detection in images. It contains 47,776 images in total, with 38,118 allocated to the training set and 9,658 to the test set. The dataset encompasses 600 HOI categories derived from 80 object categories and 117 verb categories, and provides over 150,000 annotated human-object pairs.

V-COCO includes 10,346 images: 2,533 for training, 2,867 for validation, and 4,946 for testing, along with 16,199 human instances. Each human instance is annotated with 29 action categories, and no interaction labels involving objects are included.