There are 41712 image annotations in halpe_train_v1.json, but in HICO-DET dataset, it contains 47,776 images (38,118 in train set and 9,658 in test set)

Fang-Haoshu / Halpe-FullBody

Halpe: full body human pose estimation and human-object interaction detection dataset

328 stars 32 forks source link

Open wangwei-630 opened 2 years ago

wangwei-630 commented 2 years ago

Thus, which image in HICO-DET dataset should be used corresponding to halpe_train_v1.json?

vicentowang commented 1 year ago

41712 objs anns in 38,118 images.