You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I attempted to slightly modify the body estimation part in order to fine-tune the backbone network, and noticed several typos. Eventually, I refactored almost everything and noticed significant improvements on estimation accuracy, although using the same pretrained model and thresholds. Since I modified some APIs, and I can't recall what changes I made was critical to the improvement, instead of making a pull request I would like to ask the original contributors to take a look at my version and see if something can help in improving things here.
I just found the main source of improvement. My default scale search option is [0.7, 1.0, 1.3], while the original scale_search is [0.5]. Sorry for the bothering, but I think correcting the typos would still help the readers in the long term, so I would keep the issue open.
I attempted to slightly modify the body estimation part in order to fine-tune the backbone network, and noticed several typos. Eventually, I refactored almost everything and noticed significant improvements on estimation accuracy, although using the same pretrained model and thresholds. Since I modified some APIs, and I can't recall what changes I made was critical to the improvement, instead of making a pull request I would like to ask the original contributors to take a look at my version and see if something can help in improving things here.
Some differences:
Some of the changes I remember are listed below, but I think they are not the main reason of the improvement:
util.draw_bodypose
to align with other implementations better.util.draw_bodypose
. The original version is not erroneous but just misleading._L1
to_L2
at the end of this lineThe text was updated successfully, but these errors were encountered: