Filtering keypoints to be plotted

Support
,
1

I’m using yolo11n-pose to identify keypoints in a video of a person’s face. It’s flagging keypoints for the nose, eyes, and ears, as shown in the image below:

image
image650×488 65.2 KB

I’m trying to figure out how to generate a visualization that only shows the keypoints for the eyes and nose. (#s 0, 1, and 2, I believe?)

[simplified] My code is as follows:

results = model(vid_source)
r.show(kpt_line=False)

To get at the keypoints in question, I indexed into results.keypoints.xy[0][0:3] and got the subset of results I was expecting, but I can’t figure out how to do the same for the object getting passed to the plotting utility.

2

There isn’t a direct way to accomplish this with the .plot() method. You could use a bit of a roundabout way manually

from copy import deepcopy

from ultralytics import YOLO
from ultralytics.utils.plotting import Annotator, colors


# Using all default config values
line_width = None
font = "Arial.ttf"
font_size = None
kpt_radius = 5
kpt_line = False
color_mode = "instance"

model = YOLO("yolo11n-pose.pt")

src = ""  # update as needed
results = model.predict(src)

for result in results:
    annotator = Annotator(
        deepcopy(result.orig_img),  # could use frame if iterating
        line_width,
        font_size,
        font,
        example=model.names,
    )
    for i, k in enumerate(reversed(results.keypoints.data)):
        annotator.kpts(
            k[:3],  # might have to adjust as needed
            results.orig_shape,
            radius=kpt_radius,
            kpt_line=kpt_line,
            kpt_color=colors(i, True) if color_mode == "instance" else None,
    )
    annotator.show(result.path)

Ripped out from here and you can see more details about the Annotator class here.

1 Like