def full_image_similarity(image):
image_rgb = skutil.img_as_float(image)
image_gray = cl.rgb2gray(image_rgb)
gabor_responses = feat.compute_gabor_responses(image_gray, gabor_kernels)
active_mask = np.ones(image.shape).astype('uint8') * 255
query_features = feat.sim_feature_extraction(image_rgb, gabor_responses,
active_mask)
def feature_extraction(image, mask):
"""
similarity feature extraction
"""
image_gray = cl.rgb2gray(skutil.img_as_float(image))
# get gabor response from the whole image
gabor_responses = feat.compute_gabor_responses(image_gray, gabor_kernels)
# extract features
return feat.sim_feature_extraction(image, gabor_responses, mask)
image = tran.rescale(image, scale=(scale, scale))
image_gray = color.rgb2gray(image)
# load mask
mask_path = args.image_query.split(extension)[0] + "_mask" + extension
mmask = io.imread(mask_path)
mmask = tran.rescale(mmask, scale=(scale, scale))
mask = np.zeros(mmask.shape, dtype='uint8')
mask[mmask > 0] = 255
mask = mask.reshape(mask.shape[0], mask.shape[1], 1).repeat(3, axis=2)
# get gabor response from the whole image
gabor_responses = feat.compute_gabor_responses(image_gray, gabor_kernels)
# extract features from test image
query_features = feat.sim_feature_extraction(image, gabor_responses, mask)
# build KD Tree
K = 10
dists, indices = tree.query(query_features.reshape(1, -1), k=K)
# visualize query image
fig = plt.figure("vis")
ax0 = fig.add_subplot(1, 11, 1)
ax0.imshow(image)
ax0.set_xticks([])
ax0.set_yticks([])
ax0.set_title("query image")
# visualize results
for i, idx in enumerate(indices[0]):
ax3 = fig.add_subplot(2, n_axis, i + 3)
ax3.set_title(true_labels[str(label)])
ax3.imshow(cropped)
# get similar clothings
(tree, image_paths) = pickle.load(open(args.sim_features, 'rb'))
# get gabor response from the region
image_rgb = skutil.img_as_float(image)
image_gray = cl.rgb2gray(image_rgb)
gabor_responses = feat.compute_gabor_responses(image_gray, gabor_kernels)
# extract features from test image
active_mask = segment_masks[unique_labels.index(4)].astype('uint8') * 255
query_features = feat.sim_feature_extraction(image_rgb, gabor_responses,
active_mask)
# build KD Tree
dists, indices = tree.query(query_features.reshape(1, -1), k=n_axis)
# visualize similar results
for i, idx in enumerate(indices[0]):
im = io.imread(image_paths[idx]['full'])
ax4 = fig.add_subplot(2, n_axis, i + n_axis + 1)
ax4.imshow(im)
ax4.set_xticks([])
ax4.set_yticks([])
ax4.set_title("similar #%i" % (i + 1))
plt.show()
plt.show()