def test_gradcheck(self, device):
B, C, H, W = 1, 1, 32, 32
img = torch.rand(B, C, H, W, device=device)
img = utils.tensor_to_gradcheck_var(img) # to var
local_feature = GFTTAffNetHardNet(2, True).to(device, img.dtype)
assert gradcheck(local_feature,
img,
eps=1e-4,
atol=1e-4,
raise_exception=True)
def test_nomatch(self, device, dtype, data):
matcher = LocalFeatureMatcher(GFTTAffNetHardNet(100),
DescriptorMatcher('snn', 0.8)).to(
device, dtype)
data_dev = utils.dict_to(data, device, dtype)
with torch.no_grad():
out = matcher({
"image0": data_dev["image0"],
"image1": 0 * data_dev["image0"]
})
assert len(out['keypoints0']) == 0
def test_real_gftt(self, device, dtype, data):
torch.random.manual_seed(0)
# This is not unit test, but that is quite good integration test
matcher = LocalFeatureMatcher(GFTTAffNetHardNet(2000),
DescriptorMatcher('snn', 0.8)).to(
device, dtype)
ransac = RANSAC('homography', 1.0, 2048, 10).to(device, dtype)
data_dev = utils.dict_to(data, device, dtype)
pts_src = data_dev['pts0']
pts_dst = data_dev['pts1']
with torch.no_grad():
out = matcher(data_dev)
homography, inliers = ransac(out['keypoints0'], out['keypoints1'])
assert inliers.sum().item() > 50 # we have enough inliers
# Reprojection error of 5px is OK
assert_close(transform_points(homography[None], pts_src[None]),
pts_dst[None],
rtol=5e-2,
atol=5)
def __init__(self,
initial_matcher: Optional[LocalFeature] = None,
fast_matcher: Optional[nn.Module] = None,
ransac: Optional[nn.Module] = None,
minimum_inliers_num: int = 30) -> None:
super().__init__()
self.initial_matcher = initial_matcher or (LocalFeatureMatcher(
GFTTAffNetHardNet(3000), DescriptorMatcher('smnn', 0.95)))
self.fast_matcher = fast_matcher or LoFTR('outdoor')
self.ransac = ransac or RANSAC('homography',
inl_th=5.0,
batch_size=4096,
max_iter=10,
max_lo_iters=10)
self.minimum_inliers_num = minimum_inliers_num
# placeholders
self.target: torch.Tensor
self.target_initial_representation: Dict[str, torch.Tensor] = {}
self.target_fast_representation: Dict[str, torch.Tensor] = {}
self.previous_homography: Optional[torch.Tensor] = None
self.reset_tracking()
def test_smoke(self, device, dtype):
feat = GFTTAffNetHardNet().to(device, dtype)
assert feat is not None