def test_chamfer():
distChamfer = ext.chamferDist()
p1 = torch.rand(4, 100, 3).cuda()
p2 = torch.rand(4, 200, 3).cuda()
points1 = Variable(p1, requires_grad=True)
points2 = Variable(p2)
dist1, dist2, idx1, idx2= distChamfer(points1, points2)
loss = torch.sum(dist1)
print(loss)
loss.backward()
print(points1.grad, points2.grad)
mydist1, mydist2, myidx1, myidx2 = chamfer_python.distChamfer(points1, points2)
d1 = (dist1 - mydist1) ** 2
d2 = (dist2 - mydist2) ** 2
print(d1, d2)
assert (
torch.sum(d1) + torch.sum(d2) < 0.00000001
), "chamfer cuda and chamfer normal are not giving the same results"
xd1 = idx1 - myidx1
xd2 = idx2 - myidx2
assert (
torch.norm(xd1.float()) + torch.norm(xd2.float()) == 0
), "chamfer cuda and chamfer normal are not giving the same results"
def test_high_dims():
distChamfer = ext.chamferDist()
p1 = torch.rand(4, 100, 5).cuda()
p2 = torch.rand(4, 200, 5).cuda()
points1 = Variable(p1, requires_grad=True)
points2 = Variable(p2)
mydist1, mydist2, idx1, idx2 = chamfer_python.distChamfer(points1, points2)
print(mydist1, mydist2, idx1, idx2)
def get(opt):
if opt.accelerated_chamfer:
import dist_chamfer_idx as ext
distChamfer = ext.chamferDist()
else:
import chamfer_python
distChamfer = chamfer_python.distChamfer
return distChamfer
def timings():
distChamfer = ext.chamferDist()
p1 = torch.rand(2, 5, 3).cuda()
p2 = torch.rand(2, 10, 3).cuda()
print("Start CUDA version")
start = time.time()
for i in range(1000):
points1 = Variable(p1, requires_grad=True)
points2 = Variable(p2)
mydist1, mydist2, idx1, idx2 = distChamfer(points1, points2)
print(mydist1.size(), mydist2.size(), idx1.size(), idx2.size())
print(idx1)
loss = torch.sum(mydist1)
loss.backward()
print("Ellapsed time is", {time.time() - start}, "seconds.")
def timings():
distChamfer = ext.chamferDist()
p1 = torch.rand(32, 2000, 3).cuda()
p2 = torch.rand(32, 1000, 3).cuda()
print("Start CUDA version")
start = time.time()
for i in range(1000):
points1 = Variable(p1, requires_grad=True)
points2 = Variable(p2)
mydist1, mydist2, idx1, idx2 = distChamfer(points1, points2)
loss = torch.sum(mydist1)
loss.backward()
print(f"Ellapsed time is {time.time() - start} seconds.")
print("Start Pythonic version")
start = time.time()
for i in range(1000):
points1 = Variable(p1, requires_grad=True)
points2 = Variable(p2)
mydist1, mydist2, idx1, idx2 = chamfer_python.distChamfer(points1, points2)
loss = torch.sum(mydist1)
loss.backward()
print(f"Ellapsed time is {time.time() - start} seconds.")
import torch
import dist_chamfer_idx as ext
import chamfer_python
distChamfer = ext.chamferDist()
from torch.autograd import Variable
import time
def test_chamfer():
distChamfer = ext.chamferDist()
p1 = torch.rand(4, 100, 3).cuda()
p2 = torch.rand(4, 200, 3).cuda()
points1 = Variable(p1, requires_grad=True)
points2 = Variable(p2)
dist1, dist2, idx1, idx2= distChamfer(points1, points2)
loss = torch.sum(dist1)
print(loss)
loss.backward()
print(points1.grad, points2.grad)
mydist1, mydist2, myidx1, myidx2 = chamfer_python.distChamfer(points1, points2)
d1 = (dist1 - mydist1) ** 2
d2 = (dist2 - mydist2) ** 2
print(d1, d2)
assert (
torch.sum(d1) + torch.sum(d2) < 0.00000001
), "chamfer cuda and chamfer normal are not giving the same results"
xd1 = idx1 - myidx1
xd2 = idx2 - myidx2