def prepare_full_feature():
image_base_path = "../images/kyoto/"
model_path = "/home/ge/tests/vgg16_weights.h5"
FeatureExtractor.initialize(model_path)
images = np.zeros((500, 4096))
for i in range(500):
img = imread(image_base_path + str(i) + ".jpg")
feature = FeatureExtractor.feature(img)
images[i, :] = feature
np.save("../mid-data/full_feature.npy", images)
refined_proposals = []
test_image_path = "../test_images/kyoto2/"
image_base_path = "../images/kyoto/"
model_path = "/home/ge/tests/vgg16_weights.h5"
FeatureExtractor.initialize(model_path)
for iter in range(1):
if iter == 0:
KNN.initialize("../mid-data/full_feature.npy")
else:
distance_mat = np.zeros((dataset_size, feature_length))
for i in range(dataset_size):
image = imread(image_base_path + str(i) + ".jpg")
if len(refined_proposals[0][i]) != 0:
t_refined_box = refined_proposals[0][i]
image = image[t_refined_box[0]:t_refined_box[2], t_refined_box[1]:t_refined_box[3]]
distance_mat[i, :] = FeatureExtractor.feature(image)
KNN.set_distance_mat(distance_mat)
refined_proposals = main_iterate()
print str(iter) + " th iteration finished"
refined_box = refined_proposals[0]
candidate_box = refined_proposals[1]
np.save("bbox.npy", refined_proposals)
for i in range(dataset_size):
image = imread(image_base_path + str(i) + ".jpg")
t_refined_box = refined_box[i]
if len(t_refined_box) == 0:
continue
for j in range(len(candidate_box[i])):
proposal = candidate_box[i][j]
if proposal[2] >= image.shape[1]:
