Safran/KPConv-PyTorch
1
0
Fork 0
Code Issues Pull requests Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'origin/master'

Browse source
This commit is contained in:
HuguesTHOMAS 2021-03-02 09:53:49 -05:00
parent 47591cc637 bb0ea445b1
commit 7fefb6a8d3
1 changed files with 11 additions and 3 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

14
kernels/kernel_points.py
View file

@ -318,7 +318,11 @@ def kernel_point_optimization_debug(radius, num_points, num_kernels=1, dimension
saved_gradient_norms = np.zeros((10000, num_kernels)) saved_gradient_norms = np.zeros((10000, num_kernels))
old_gradient_norms = np.zeros((num_kernels, num_points)) old_gradient_norms = np.zeros((num_kernels, num_points))
for iter in range(10000): step = -1
while step < 10000:
# Increment
step += 1
# Compute gradients # Compute gradients
# ***************** # *****************
@ -344,7 +348,7 @@ def kernel_point_optimization_debug(radius, num_points, num_kernels=1, dimension
# Compute norm of gradients # Compute norm of gradients
gradients_norms = np.sqrt(np.sum(np.power(gradients, 2), axis=-1)) gradients_norms = np.sqrt(np.sum(np.power(gradients, 2), axis=-1))
saved_gradient_norms[iter, :] = np.max(gradients_norms, axis=1) saved_gradient_norms[step, :] = np.max(gradients_norms, axis=1)
# Stop if all moving points are gradients fixed (low gradients diff) # Stop if all moving points are gradients fixed (low gradients diff)
@ -372,7 +376,7 @@ def kernel_point_optimization_debug(radius, num_points, num_kernels=1, dimension
kernel_points -= np.expand_dims(moving_dists, -1) * gradients / np.expand_dims(gradients_norms + 1e-6, -1) kernel_points -= np.expand_dims(moving_dists, -1) * gradients / np.expand_dims(gradients_norms + 1e-6, -1)
if verbose: if verbose:
print('iter {:5d} / max grad = {:f}'.format(iter, np.max(gradients_norms[:, 3:]))) print('step {:5d} / max grad = {:f}'.format(step, np.max(gradients_norms[:, 3:])))
if verbose > 1: if verbose > 1:
plt.clf() plt.clf()
plt.plot(kernel_points[0, :, 0], kernel_points[0, :, 1], '.') plt.plot(kernel_points[0, :, 0], kernel_points[0, :, 1], '.')
@ -389,6 +393,10 @@ def kernel_point_optimization_debug(radius, num_points, num_kernels=1, dimension
# moving factor decay # moving factor decay
moving_factor *= continuous_moving_decay moving_factor *= continuous_moving_decay
# Remove unused lines in the saved gradients
if step < 10000:
saved_gradient_norms = saved_gradient_norms[:step+1, :]
# Rescale radius to fit the wanted ratio of radius # Rescale radius to fit the wanted ratio of radius
r = np.sqrt(np.sum(np.power(kernel_points, 2), axis=-1)) r = np.sqrt(np.sum(np.power(kernel_points, 2), axis=-1))
kernel_points *= ratio / np.mean(r[:, 1:]) kernel_points *= ratio / np.mean(r[:, 1:])