Pose model failure reason

[qhaotian0525]

Have I written custom code (as opposed to using a stock example script provided in MediaPipe)

No

OS Platform and Distribution

Windows 11

MediaPipe version

latest

Bazel version

No response

Solution

Pose

Programming Language and version

Python

Describe the actual behavior

It failed to detect the second person in some images but the person is clear to view.

Describe the expected behaviour

It should detect both of them.

Standalone code/steps you may have used to try to get what you need

https://github.com/qhaotian0525/posefailcases.git

This is the link to the repo. It contains several images from the dataset I created. The pretrained pose model failed partially on most of them. It can detect one person only. I noticed that if two persons are on the same plane/distance, it's more likely to detect successfully. If they are of different scales (different distance), it will only detect the larger person. I set `num_poses = 5`, so I think the maximum number should be 5.

Did I miss anything in the model configuration causing this low detection rate? Any way to improve the results?

Thank you!

Other info / Complete Logs

No response