Why create a "global_part_neighbor" in the function "find_neighbor" in the skeleton.py?

[ANYMS-A]

Hi,
I was recently reading the source code very carefully in order to understand the whole pipeline of your work.

I was kind of confused that why init a variable named "global_part_neighbor" in the in the function "find_neighbor" in the line #373 of the models/skeleton.py?

Besides, why append a "edge_num" into those global_part_neighbor's list in line #375?

Many thanks!

[PeizhuoLi]

Hi,

Line 373 is because we treat the global position channel as an "extra joint" and it shares the neighbor of the root joint, which is exactly global_part_neighbor.

For line 375, the "neighbor" relation is symmetric, so we add the extra joint (its index is edge_num, the last one) to the neighbor_list of its neighbors.

Hope this will help.

[ANYMS-A]

Hi,

Line 373 is because we treat the global position channel as an "extra joint" and it shares the neighbor of the root joint, which is exactly global_part_neighbor.

For line 375, the "neighbor" relation is symmetric, so we add the extra joint (its index is edge_num, the last one) to the neighbor_list of its neighbors.

Hope this will help.

Got it !Thank you so much for the answer!

[ANYMS-A]

Hi, I recently found that there might be a mistake when calculating the neighbors of the "global_part_neighbor".

e.g. If you read the .bvh file of "BigVegas", you can use debug mode of the IDE to found that the 22th element of the list which returned by the "find_neighbor" function is not the neighbors of the root joint(if you draw a graph of these 23 bones). It should be [0, 1, 4, 5, 8, 9], but got [0, 1, 2, 4, 5, 8, 9] instead.

I thinks it is because: at first, the index of the root joint is 0, but the "edges" eliminate the root joint, so when we obtain the neighbor_list after the for loop in line #370 of the skeleton.py, the mistake occurs.

global_part_neighbor = neighbor_list[0].copy()

is not the neighbors of the root joint. I think it represents the neighbors of the "LeftUpLeg" of the "BigVegas"'s skeleton.(you could draw out the skeleton graph of the "BigVegas"'s 23 bones to verify it.)

[PeizhuoLi]

Thank you very mush for providing your findings! I think it's indeed a bug and we'll fix it later.

[fire]

Can you describe a way for fixing this? I can retrain my models.

[PeizhuoLi]

global_part_neighbor should be all the edges(armature) that contains joint with index zero. This condition should be used instead of a copy of neighbor_list[0].

[jdbodyfelt]

So if neighbor_list = [] how can this be fixed?

[fire]

@PeizhuoLi I'm having trouble with neighbor_list = []

[PeizhuoLi]

Could you provide more details on how to reproduce this error? Thanks :)