how to get keypoint results from dino-x API

[sh2120170809]

Using demo.py, I can get detection boxes and masks, but I am trying to get keypoints for person and hand. How can I get keypoint results?

[rentainhe]

Hello, we will refine our documentation recently and add keypoints demo

[reb1rth-l]

Hello，I want to get keypoints for person and hands too，it's two months since your last replay, I want to know when will it release?

Hello, we will refine our documentation recently and add keypoints demo

[rentainhe]

Hello, sorry for the late reply, you can get the keypoints annotations by updating your API usage as:

task = DinoxTask(
    image_url=image_url,
    prompts=[TextPrompt(text=TEXT_PROMPT)],
    bbox_threshold=0.25,
    targets=[DetectionTarget.BBox, DetectionTarget.Mask, DetectionTarget.Pose]  # add Pose target
)

and the prediction results will contain the key points predictons in this format (x, y, visibility, _)

• Notes: it works for the text prompts relative with person, e.g. person, teacher, man.

We will update the visualization in two days.