datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
hico-det-llava-v1.6-13b-answers
HICO-DET LLaVA-1.6-13B action answers
Per-image action lists for every HICO-DET image (38,118 train / 9,658 test),
produced by LLaVA-1.6 (vicuna-13B) prompted with the 117 HICO-DET verb
names and asked to list at most 7 valid actions visible in the picture.
These are the text-side VLM answers consumed by UMI-HOI (Unified
Multimodal Interaction HOI detection) at training and test time.
No images are included; obtain HICO-DET separately and join on image.
Files… See the full description on the dataset page: https://huggingface.co/datasets/Pikaqiu0114/hico-det-llava-v1.6-13b-answers.hico-clip-det