V-COCO Human-Object Interaction Detection Dataset

V-COCO stands for Verbs in COCO. It is a dataset based on the MS COCO dataset and is used for interaction detection between people and objects.

The dataset provides 10,346 images (2,533 for training, 2,867 for validation, and 4,946 for testing) and 16,199 person instances. Each person is annotated with 29 action categories. The dataset does not have interaction labels including objects.