Affordance data issue

[roammonth]

Hello.
I hope this time I can receive a response from you.
The description of the contact data obtained in the get_affordance.py file seems to be inconsistent with the paper. In the paper, the contact between 8 main joints and an object is described, while in the code, the contact between 22 joints and an object is calculated.
In addition, in many sequences where there is contact between the human body and objects, why are all the contact data results obtained zero.

[xiaogangpeng]

Hello, thanks for following up — and sorry for the earlier confusion.

We’ve now fixed the issue so that the contact extraction strictly follows the paper. Contact is computed only for the 8 primary joints, and for each joint we select the most stable / most frequently occurring contact point over the entire sequence.

In addition, the cases where some sequences produced all-zero contact data were caused by the previous distance thresholding between the object surface and the human joints. This has been corrected by first computing marker-to-object distances for human–object contact, and then performing marker-to-joint mapping, which produces more accurate and consistent contact annotations.

Please feel free to test the updated version, and let me know if anything else looks unexpected. Thank you again for pointing this out!