How did you filter out some of the detected skeletons?

[chamecall]

Hello again) I decided to use some other pose estimation detector on video from Avenue dataset (01 testing video). I've obtained many more skeletons than you have in your ready masks.

So as an example for the same 01 testing video for the first frame I've obtained the following poses

while you have only one pose detection for that frame across all the csv masks

so the questions are:

• did you filter out the skeletons on purpose?
• what criteria have you used for obtaining that?

I'm asking just because I get abnormal scores for some of these skeletons that aren't present among your masks. Thanks in advance.

[RomeroBarata]

Hi chamecall,

If I remember correctly, the only criterion I used to filter out skeletons was if their trajectory was too short. This is the method to filter and some of its use here.

Kind regards, Romero