parser.set_defaults(log=False)
args = parser.parse_args()
print('Config path: {}'.format(args.config))
print('Model config path: {}'.format(args.model_config))
if args.weights is None:
w = 'RANDOM WEIGHTS'
else:
w = args.weights
print('Weights: {}'.format(w))
print('Debug mode: {}'.format(args.debug))
print('Visualize: {}'.format(args.visualize))
print('Log search results: {}'.format(args.log))
print('')
params = MinkLocParams(args.config, args.model_config)
params.print()
if torch.cuda.is_available():
device = "cuda"
else:
device = "cpu"
print('Device: {}'.format(device))
model = model_factory(params)
if args.weights is not None:
assert os.path.exists(
args.weights), 'Cannot open network weights: {}'.format(
args.weights)
print('Loading weights: {}'.format(args.weights))
model.load_state_dict(torch.load(args.weights, map_location=device))
# Default downsampling factor 1280x960 images to 320x240
parser.add_argument('--downsample', type=int, default=4, help='Image downsampling factor')
args = parser.parse_args()
print('Config path: {}'.format(args.config))
print('Oxford RobotCar root folder: {}'.format(args.oxford_root))
print('Camera: {}'.format(args.camera))
print('Image downsampling factor: {}'.format(args.downsample))
nn_threshold = 1000 # Nearest neighbour threshold in miliseconds
k = 20 # Number of nearest neighbour images to find for each LiDAR scan
ext = '.png' # Image extension
print('Number of nearest images for each scan (k): {}'.format(k))
print('')
params = MinkLocParams(args.config, model_params_path=None)
print(f'Parameters from config file: {args.config}')
print(f"Output folder for downsampled images (image_path): {params.image_path}")
print(f"Dataset folder (point clouds): {params.dataset_folder}")
print(f"Evaluation sets - query split: {params.eval_query_files}")
print(f"Evaluation sets - database split: {params.eval_database_files}")
print('')
# Create output path
out_path = params.image_path
if not os.path.exists(out_path):
os.mkdir(out_path)
assert os.path.exists(out_path), 'Cannot create output directory: {}'.format(out_path)
# Index LiDAR scans in the dataset