tried running mediapose on some examples from leed sports dataset

[Sparsh Garg]

I tried running mediapipe pose estimation using colab on some images from Leed sports dataset.These are the results I got.

Based on what is described under holistic pose esitmation on the mediapipe github page and other pose estimation algorithms like OpenPose,there is a huge discrepancy here.

Any ideas as to what am I doing wrong here

is google colab pose estimation a good way to try out the library,or should I tune the parameters.If I need to tune the parameters,then any suggestions on how to do that will be appreciated.