Prior to adding the neighbor statements (that I commented with 'new'), everything worked just fine. Now, when using numpy.asarray, there is following error:
ValueError: could not broadcast input array from shape (3,3) into shape (3).
I'm really confused as the new lines didn't change anything about the rotations-array.
def rre(mesh, rotations):
"""
Relative Rotation Encoding (RRE).
Return a compact representation of all relative face rotations.
"""
all_rel_rotations = neighbors = []
for f in mesh.faces():
temp = [] # new
for n in mesh.ff(f):
rel_rotation = np.matmul(rotations[f.idx()], np.linalg.inv(rotations[n.idx()]))
all_rel_rotations.append(rel_rotation)
temp.append(n.idx()) # new
neighbors.append(temp) # new
all_rel_rotations = np.asarray(all_rel_rotations)
neighbors = np.asarray(neighbors) # new
return all_rel_rotations, neighbors
The source of the problem is very likely the line:
all_rel_rotations = neighbors = []
In Python lists are mutables and all_rel_rotations and neighbors point to the same list, so if you do all_rel_rotations.append(42) you will see that neighbors = [42, ]
The line:
all_rel_rotations.append(rel_rotation)
appends a 2D array, while
neighbors.append(temp)
appends a 1D array (or the other way around) to the same list. Then:
all_rel_rotations = np.asarray(all_rel_rotations)
tries to convert to an array and get confused.
If you need to list do
all_rel_rotations = []
neighbors = []