jacklangerman commited on
Commit
91b350c
1 Parent(s): a842b44

refactor cv + fix calc

Browse files
Files changed (1) hide show
  1. hoho/wed.py +12 -10
hoho/wed.py CHANGED
@@ -28,13 +28,13 @@ def preregister_mean_std(verts_to_transform, target_verts, single_scale=True):
28
  return transformed_verts
29
 
30
 
31
- def compute_WED(pd_vertices, pd_edges, gt_vertices, gt_edges, cv=-1, ce=1.0, normalized=True, preregister=True, single_scale=True):
32
  '''The function computes the Wireframe Edge Distance (WED) between two graphs.
33
  pd_vertices: list of predicted vertices
34
  pd_edges: list of predicted edges
35
  gt_vertices: list of ground truth vertices
36
  gt_edges: list of ground truth edges
37
- cv: vertex cost (the cost in centimeters of missing a vertex, default is -1, which means 1/4 of the diameter of the ground truth mesh)
38
  ce: edge cost (multiplier of the edge length for edge deletion and insertion, default is 1.0)
39
  normalized: if True, the WED is normalized by the total length of the ground truth edges
40
  preregister: if True, the predicted vertices have their mean and scale matched to the ground truth vertices
@@ -43,21 +43,23 @@ def compute_WED(pd_vertices, pd_edges, gt_vertices, gt_edges, cv=-1, ce=1.0, nor
43
  # Vertex coordinates are in centimeters. When cv and ce are set to 100.0 and 1.0 respectively,
44
  # missing a vertex is equivanlent predicting it 1 meter away from the ground truth vertex.
45
  # This is equivalent to setting cv=1 and ce=1 when the vertex coordinates are in meters.
46
- # When a negative cv value is set (the default behavior), cv is reset to 1/4 of the diameter of the ground truth wireframe.
47
 
48
  pd_vertices = np.array(pd_vertices)
49
  gt_vertices = np.array(gt_vertices)
50
 
51
- diameter = cdist(gt_vertices, gt_vertices).max()
52
-
53
- if cv < 0:
54
- cv = diameter / 4.0
55
- # Cost of addining or deleting a vertex is set to 1/4 of the diameter of the ground truth mesh
56
-
57
- # Step 0: Prenormalize / preregister
58
  if preregister:
59
  pd_vertices = preregister_mean_std(pd_vertices, gt_vertices, single_scale=single_scale)
60
 
 
 
 
 
 
 
 
 
 
61
 
62
  pd_edges = np.array(pd_edges)
63
  gt_edges = np.array(gt_edges)
 
28
  return transformed_verts
29
 
30
 
31
+ def compute_WED(pd_vertices, pd_edges, gt_vertices, gt_edges, cv=-1/4, ce=1.0, normalized=True, preregister=True, single_scale=True):
32
  '''The function computes the Wireframe Edge Distance (WED) between two graphs.
33
  pd_vertices: list of predicted vertices
34
  pd_edges: list of predicted edges
35
  gt_vertices: list of ground truth vertices
36
  gt_edges: list of ground truth edges
37
+ cv: vertex cost: if positive, the cost in centimeters of missing a vertex, if negative, multiplies diameter to compute cost (default is -1/2)
38
  ce: edge cost (multiplier of the edge length for edge deletion and insertion, default is 1.0)
39
  normalized: if True, the WED is normalized by the total length of the ground truth edges
40
  preregister: if True, the predicted vertices have their mean and scale matched to the ground truth vertices
 
43
  # Vertex coordinates are in centimeters. When cv and ce are set to 100.0 and 1.0 respectively,
44
  # missing a vertex is equivanlent predicting it 1 meter away from the ground truth vertex.
45
  # This is equivalent to setting cv=1 and ce=1 when the vertex coordinates are in meters.
46
+ # When a negative cv value is set (the default behavior), cv is reset to 1/2 of the diameter of the ground truth wireframe.
47
 
48
  pd_vertices = np.array(pd_vertices)
49
  gt_vertices = np.array(gt_vertices)
50
 
 
 
 
 
 
 
 
51
  if preregister:
52
  pd_vertices = preregister_mean_std(pd_vertices, gt_vertices, single_scale=single_scale)
53
 
54
+ if cv < 0:
55
+ diameter = cdist(gt_vertices, gt_vertices).max()
56
+ # Cost of adding or deleting a vertex is set to -cv times the diameter of the ground truth wireframe
57
+ cv = -cv * diameter
58
+ elif cv == 0:
59
+ # Cost of adding or deleting a vertex is set to the average distance of the ground truth vertices from their mean
60
+ cv = np.linalg.norm(np.mean(gt_vertices, axis=0) - gt_vertices, axis=1).mean()
61
+ # Step 0: Prenormalize / preregister
62
+
63
 
64
  pd_edges = np.array(pd_edges)
65
  gt_edges = np.array(gt_edges)