updates

src/metrics.py

@@ -30,7 +30,7 @@ def weight(p, nbhd, kernel='rbf', sigma=None):
     if kernel == 'rbf':  # Gaussian
         w = np.exp(-dist ** 2 / (2 * sigma ** 2))
     elif kernel == 'cosine':
-        w = (X @ p) / np.linalg.norm(nbhd * p, axis=1)
+        w = (nbhd @ p) / np.linalg.norm(nbhd * p, axis=1)
     elif kernel == 'linear':
         w = np.maximum(1 - sigma * dist, 0)
     elif kernel == 'inverse':

src/orientation.py

@@ -16,7 +16,7 @@ def orient_normals_cvx(xyz, n):
     ----------
     xyz : numpy.ndarray
         The point cloud of shape (N, 3), N is the number of points.
-    n : numpy.ndsarray
+    n : numpy.ndarray
         The unit normals of shape (N, 3), where N is the number of
         points in the point cloud.
     
@@ -77,7 +77,6 @@ def _compute_kgraph(xyz, k):
     ----------
     xyz : numpy.ndarray
         The point cloud of shape (N, 3), N is the number of points.
-    n : numpy.ndsarray
     k : int
         Number of k nearest neighbors used in constructing the
         Riemannian graph used to propagate normal orientation.
@@ -98,7 +97,7 @@ def _compute_rmst(xyz, n, k, eps=1e-4):
     ----------
     xyz : numpy.ndarray
         The point cloud of shape (N, 3), N is the number of points.
-    n : numpy.ndsarray
+    n : numpy.ndarray
         The unit normals of shape (N, 3), where N is the number of
         points in the point cloud.
     k : int

src/utils.py

@@ -51,4 +51,4 @@ def polyfit2d(x, y, z, deg=1, rcond=None, full_output=False):
     coef, res, rank, s = np.linalg.lstsq(A, b, rcond=rcond)
     if full_output:
         return coef.reshape(deg+1, deg+1), res, rank, s
-    return coef.reshape(deg+1, deg+1)
+    return coef.reshape(deg+1, deg+1)