calculate the pixel length in meters

[MasterMedo]

Based on the subject height and body part length ratios, calculate the length of n pixel in meters.

average_height = 1.75  # height in meters

def pixel_meter_ratio(body_part_positions: list[list[float, float, float]], height: float = average_height) -> float

Takes a list of body part positions (17 body parts). The elements are [x coordinate in pixels, y coordinate in pixels, confidence]:

[
    [320, 524, 0.123],
    [320, 524, 0.123],
    [320, 524, 0.123],
    [320, 524, 0.123],
    [320, 524, 0.123],
    [320, 524, 0.123],
    [320, 524, 0.123],
    [320, 524, 0.123],
    [320, 524, 0.123],
    [320, 524, 0.123],
    [320, 524, 0.123],
    [320, 524, 0.123],
    [320, 524, 0.123],
    [320, 524, 0.123],
    [320, 524, 0.123],
    [320, 524, 0.123],
    [320, 524, 0.123],
]

Returns the pixel/meter ratio:

106.13

[MasterMedo]

The code we have currently for calculating the ratio of a pixel and a meter works decently when the athlete is facing the camera or is turned away from the camera. It also works great if the camera sees the athlete's profile if the torso or the thighs can be detected.

What currently doesn't work so well is when the athlete's profile waist up is the only thing the camera sees. Then the algorithm concludes falsely that the athlete is facing the camera. It concludes the person is far away from the camera and tiny, which isn't the case. I propose assuming we can see more of the athlete's body the farther the athlete is from the camera. Other suggestions are welcome.