def load_img(config, data_dir):
img, geom, vis, depth, kp, desc = loadFromDir(
data_dir,
"-16x16",
bUseColorImage=True,
crop_center=config.data_crop_center,
load_lift=config.use_lift)
return img, geom, vis
# train_path = getattr(config, "data_dir_" + _set[:2]) + split + "/"
train_path = getattr(config, "data_dir_tr")
# Create data dump directory name
data_names = getattr(config, "data_tr")
data_name = data_names.split(".")[0]
cur_folder = "/".join(
[data_folder, data_name, "numkp-{}".format(config.obj_num_kp)])
if not os.path.exists(cur_folder):
os.makedirs(cur_folder)
img, kp, desc, aff, K, R, t = loadFromDir(train_path,
cur_folder,
"-16x16",
bUseColorImage=True,
crop_center=crop_center,
load_hessian=True)
if len(kp) == 0:
kp = [None] * len(img)
if len(desc) == 0:
desc = [None] * len(img)
pair_index = np.loadtxt(train_path + "pair_index.txt")
# Check if we've done this folder already.
print(" -- Waiting for the data_folder to be ready")
ready_file = os.path.join(cur_folder, "ready")
if not os.path.exists(ready_file):
print(" -- No ready file {}".format(ready_file))
# Now start data prep
print("Preparing data for {}".format(config.data_tr.split(".")[0]))
for _set in ["train", "valid", "test"]:
num_sample = getattr(
config, "train_max_{}_sample".format(_set[:2]))
# Load the data
print("Loading Raw Data for {}".format(_set))
if _set == "valid":
split = "val"
else:
split = _set
img, geom, vis, depth, kp, desc = loadFromDir(
getattr(config, "data_dir_" + _set[:2]) + split + "/",
"-16x16",
bUseColorImage=True,
crop_center=crop_center,
load_lift=config.use_lift)
if len(kp) == 0:
kp = [None] * len(img)
if len(desc) == 0:
desc = [None] * len(img)
z = [None] * len(img)
# Generating all possible pairs
print("Generating list of all possible pairs for {}".format(_set))
pairs = []
for ii, jj in itertools.product(xrange(len(img)), xrange(len(img))):
if ii != jj:
if vis[ii][jj] > getattr(config, "data_vis_th_" + _set[:2]):
pairs.append((ii, jj))
# Now start data prep
print("Preparing data for {}".format(config.data_tr.split(".")[0]))
for _set in ["train", "valid", "test"]:
num_sample = getattr(config, "train_max_{}_sample".format(_set[:2]))
# Load the data
print("Loading Raw Data for {}".format(_set))
if _set == "valid":
split = "val"
else:
split = _set
img, geom, vis, depth, kp, desc = loadFromDir(
getattr(config, "data_dir_" + _set[:2]) + split + "/",
"-16x16",
bUseColorImage=True,
crop_center=crop_center,
precomputed_kp_method=config.precomputed_kp_method)
if len(kp) == 0:
kp = [None] * len(img)
if len(desc) == 0:
desc = [None] * len(img)
z = [None] * len(img)
# Generating all possible pairs
print("Generating list of all possible pairs for {}".format(_set))
pairs = []
for ii, jj in itertools.product(xrange(len(img)), xrange(len(img))):
if ii != jj:
if vis[ii][jj] > getattr(config, "data_vis_th_" + _set[:2]):
pairs.append((ii, jj))