def compute_dist(self, images, labels, sample_size=None):
"""Compute distances for pairs
sample_size: None or integer, number of pairs as all pairs can be large number.
If None, by default all possible pairs are considered.
Returns:
dist_embed: array of distances
actual_issame: array of booleans, True if positive pair, False if negative pair
"""
# Run forward pass to calculate embeddings
print('Generating pairs and computing embeddings...')
# Define generator to loop over data
gen = ImageDataGenerator(
data_format=K.image_data_format(),
preprocessing_function=self.backend_class.normalize,
)
features_shape = self.model.get_output_shape_at(0)[1:]
if sample_size is None:
n_pairs = comb(images.shape[0], 2, exact=True)
else:
n_pairs = int(sample_size)
embeddings = np.zeros(shape=(2, n_pairs) + features_shape)
actual_issame = np.full(shape=(n_pairs, ),
fill_value=True,
dtype=np.bool)
# Create all possible combinations of images (no repeats)
idx = 0
for i in range(images.shape[0]):
for j in range(i):
img_1 = gen.random_transform(images[i].astype(K.floatx()))
img_1 = gen.preprocessing_function(img_1)
img_2 = gen.random_transform(images[j].astype(K.floatx()))
img_2 = gen.preprocessing_function(img_2)
embeddings[0, idx] = self.model.predict_on_batch(
np.expand_dims(img_1, 0))
embeddings[1, idx] = self.model.predict_on_batch(
np.expand_dims(img_2, 0))
if labels[i] != labels[j]:
actual_issame[idx] = False
idx += 1
if idx >= n_pairs:
break
if idx >= n_pairs:
break
print('Number of pairs in evaluation {}, number of positive {}'.format(
len(actual_issame), np.sum(actual_issame)))
dist_emb = distance(embeddings[0], embeddings[1], distance_metric=0)
return dist_emb, actual_issame
def getDistances(vectors):
nSet = len(vectors)
dists = [[] for k in range(nSet)]
for i in range(0, nSet):
for j in range(i + 1, nSet):
dists[i][j] = distance(vectors[i], vectors[j])
dists[j][i] = dists[i][j]
return dists
def _calc_dist_between_pts(self):
distances = []
points = self.shape_points
for ind in range(len(points) - 1):
current_pt = (points[ind][0], points[ind][1])
next_pt = (points[ind + 1][0], points[ind + 1][1])
distances.append(metrics.distance(next_pt, current_pt))
self.dist_between_pts = distances
def main(img_path='../data/test/017029558.jpg',
imgs_path='../data/test/',
csv_path='data.csv',
metric='0'):
# ../data/test/000316731.jpg
repo = Repository(csv_path)
feats_map = []
for fpath, _, feats in repo:
if fpath.split('/')[-1] == img_path.split('/')[-1]:
source_feats = feats
else:
feats_map.append((fpath, feats))
if metric.startswith('m'):
dim = int(metric.split('_')[-1])
all_feats = [feats[:dim] for _, feats in feats_map[:dim]]
cov = np.cov(np.array(all_feats))
VI = np.linalg.pinv(cov)
distance_map = map(
lambda (fpath, feats):
(mahalanobis(feats[:dim], source_feats[:dim], VI), fpath),
feats_map)
else:
metric = int(metric)
distance_map = map(
lambda (fpath, feats):
(distance(feats, source_feats, metric), fpath), feats_map)
distance_map.sort()
f, axarr = plt.subplots(4, 5, figsize=(20, 16))
axarr[0, 2].imshow(mpimg.imread(img_path))
axarr[0, 2].set_title("Source image", fontsize=24)
axarr[0, 2].xaxis.set_visible(False)
axarr[0, 2].yaxis.set_visible(False)
for i in [0, 1, 3, 4]:
axarr[0, i].axis('off')
for i in xrange(15):
axarr[1 + i / 5, i % 5].imshow(
mpimg.imread(os.path.join(imgs_path, distance_map[i][1])))
axarr[1 + i / 5, i % 5].set_title("{0:.3f}".format(distance_map[i][0]),
fontsize=20)
axarr[1 + i / 5, i % 5].xaxis.set_visible(False)
axarr[1 + i / 5, i % 5].yaxis.set_visible(False)
plt.tight_layout()
plt.show()
plt.close(f)