def __init__(self, min_diff=.5, **kw):
super(DecisionTree, self).__init__([('opencv', 'gray', 8)],
min_diff=min_diff,
**kw)
self._surf = impoint.SURF()
self._feat = imfeat.Histogram('gray', num_bins=4, norm=False)
self.rfc = None
def test_name(self):
image = cv2.imread('test_images/lena.jpg')
b = impoint.SURF()
s = 1
# Compute clusters and output
clusters = np.asfarray(imfeat.BoVW.cluster([image], b.compute_dense,
8))
f = imfeat.BoVW(lambda x: b.make_feature_mask(x, clusters), 8, 3)
out = f(image)
print(out.tolist())
print(out.shape)
def camera_map(features):
med_dists = []
for frame_feat0, frame_feat1 in zip(features['frame_features'],
features['frame_features'][1:]):
surf0, surf1 = frame_feat0['surf'], frame_feat1['surf']
surf_inst = impoint.SURF()
matches = surf_inst.match(surf0, surf1)
if 30 < len(matches) and 10 < min(len(surf0), len(surf1)):
pairs = [
np.asfarray([
surf0[x]['x'] - surf1[y]['x'],
surf0[x]['y'] - surf1[y]['y']
]) for x, y in matches
]
dists = [np.linalg.norm(x) for x in pairs]
med_dists.append(np.nan_to_num(np.median(dists)))
sys.stderr.write('Med_dists: %d\n' % len(med_dists))
return {'med_dists': med_dists}
def __init__(self, max_diff=5, **kw):
super(SURF, self).__init__(max_diff=max_diff, **kw)
self._surf = impoint.SURF()
def _dense_surf():
import impoint
s = impoint.SURF()
return s.compute_dense