We have released image-level and instance-level part-state annotations on HICO and HICO-DET. HOI recognition task can be modeled as a multi-label classification problem with 600 HOI categories. Given a still image, the model should tell the involved HOI categories in this image. Meanwhile, HOI detection needs to locate the interactive human and object and classify the action simultaneously.
We have labeled all the 38,116 images in train set and 9,658 images in test set of HICO dataset. For a better understanding of HOI recognition and detection tasks, you can refer to these works: HICO, Pair-wise; HICO-DET, TIN.
The annotation files can be found here: HAKE-Data.
The corresponding code and models are released in HAKE-Action.
Subsequently, instance-level part-state annotations on videos and the corresponding models will be made public soon.