extractor = FeatureExtractor(cache_folder=model_cache_folder,
which_net='vgg16',
which_layer=VC['layer'],
which_snapshot=0)
res = np.zeros((featDim, 0))
loc_set = np.zeros((5, 0))
# for ii,iid in enumerate(range(img_num)):
for ii, iid in enumerate(subset_ls):
img = cv2.imread(os.path.join(Dict['file_dir'], image_path[iid]))
# img = cv2.resize(img, (scale_size, scale_size))
img = myresize(img, scale_size, 'short')
tmp = extractor.extract_feature_image(img)[0]
assert (tmp.shape[2] == featDim)
height, width = tmp.shape[0:2]
tmp = tmp[offset:height - offset, offset:width - offset, :]
ntmp = np.transpose(tmp, (2, 0, 1))
gtmp = ntmp.reshape(ntmp.shape[0], -1)
if gtmp.shape[1] >= samp_size:
rand_idx = np.random.permutation(gtmp.shape[1])[:samp_size]
else:
rand_idx = np.random.permutation(gtmp.shape[1])[:samp_size -
gtmp.shape[1]]
rand_idx = np.append(range(gtmp.shape[1]), rand_idx)
res = np.column_stack((res, deepcopy(gtmp[:, rand_idx])))
for rr in rand_idx:
ihi, iwi = np.unravel_index(rr,
