Detection results, mAP calculation about upt HOT 2 CLOSED

Hi @yaoyaosanqi,

In UPT, we sample up to 15 humans and 15 objects in each image, which amounts to at most (15-1) * 15 = 210 pairs. That should be sufficient for majority of the images. Although there might be some niche images with a particularly large number of positive pairs, it's rare enough to just ignore.

If 114 pieces of information are considered the final detection result, then there is no way to know exactly what kind of interaction it is, and it seems that you need to specify all actions, like the visualization of actions in your reasoning.

The network produces results for all potential interactions in one go. The option to specify actions is only implemented to make the visualisations a bit neater. If you want all the results, just remove the --action flag. All results will then be printed.

Fred.

from upt.

Thank you from the bottom of my heart!

from upt.