How to show pose label in result image?

[blackCmd]

How to show pose label like this.

And If it can, Can I train pose? not object.
For example, MMdetction will show three person labels in the above image.
But I want to show another three labels in the above image like sitting, hitting and standing.

[innerlee]

And If it can, Can I train pose? not object.

Could you elaborate a little bit? is "train" a typo?

[jin-s13]

Hi. The COCO dataset does not provide action label (such as standing, hitting and sitting). You may need to train an action classifier using the human poses as input, and outputs the action type.

You can also choose to directly use the action recognition dataset. As far as I am concerned, some datasets already provide both human poses and action labels. see http:https://humaninevents.org/ for example.

[blackCmd]

Hi. The COCO dataset does not provide action label (such as standing, hitting and sitting). You may need to train an action classifier using the human poses as input, and outputs the action type.

You can also choose to directly use the action recognition dataset. As far as I am concerned, some datasets already provide both human poses and action labels. see http:https://humaninevents.org/ for example.

Oh, I see.
Then I think MMpose can inference human's action using trained keypoints not bbox. Is it right?
If you know where should I edit code for showing action label in a result image, plz help me.

[innerlee]

AD:
for rgb based action recognition: https://github.com/open-mmlab/mmaction2
for skeleton based action recognition: https://github.com/open-mmlab/mmskeleton

[blackCmd]

AD:
for rgb based action recognition: https://github.com/open-mmlab/mmaction2
for skeleton based action recognition: https://github.com/open-mmlab/mmskeleton

Unbelievable !!
It is what i want to do. Thank you : )