def main():
""" Program main entry point.
"""
args = parse_args()
devkit = kitti.Devkit(args.kitti_dir)
data_source = devkit.create_data_source(args.sequence,
kitti.OBJECT_CLASSES_PEDESTRIANS,
min_confidence=args.min_confidence)
with open(args.observation_cost_model, "rb") as f:
observation_cost_model = pickle.load(f)
with open(args.transition_cost_model, "rb") as f:
transition_cost_model = pickle.load(f)
tracker = min_cost_flow_tracker.MinCostFlowTracker(
args.entry_exit_cost,
observation_cost_model,
transition_cost_model,
args.max_num_misses,
args.miss_rate,
args.cnn_model,
optimizer_window_len=args.optimizer_window_len)
pymot_adapter = min_cost_flow_pymot.PymotAdapter(tracker)
visualization = pymotutils.MonoVisualization(
update_ms=kitti.CAMERA_UPDATE_IN_MS,
window_shape=kitti.CAMERA_IMAGE_SHAPE,
online_tracking_visualization=draw_online_tracking_results)
application = pymotutils.Application(data_source)
visualization.enable_videowriter("/tmp/detections.avi")
application.process_data(pymot_adapter, visualization)
application.compute_trajectories(interpolation=True)
visualization.enable_videowriter("/tmp/trajectories.avi")
application.play_hypotheses(visualization)
def main():
""" Program main entry point.
"""
args = parse_args()
devkit = motchallenge.Devkit(args.mot_dir, args.detection_dir)
data_source = devkit.create_data_source(args.sequence)
data_source.apply_nonmaxima_suppression(max_bbox_overlap=0.5)
with open(args.observation_cost_model, "rb") as f:
observation_cost_model = pickle.load(f)
with open(args.transition_cost_model, "rb") as f:
transition_cost_model = pickle.load(f)
tracker = min_cost_flow_tracker.MinCostFlowTracker(
args.entry_exit_cost,
observation_cost_model,
transition_cost_model,
args.max_num_misses,
args.miss_rate,
args.cnn_model,
optimizer_window_len=args.optimizer_window_len,
observation_cost_bias=args.observation_cost_bias)
pymot_adapter = min_cost_flow_pymot.PymotAdapter(tracker)
# Compute a suitable window shape.
image_shape = data_source.peek_image_shape()[::-1]
aspect_ratio = float(image_shape[0]) / image_shape[1]
window_shape = int(aspect_ratio * 600), 600
visualization = pymotutils.MonoVisualization(
update_ms=25,
window_shape=window_shape,
online_tracking_visualization=draw_online_tracking_results)
application = pymotutils.Application(data_source)
application.process_data(pymot_adapter, visualization)
application.compute_trajectories(interpolation=True)
if args.show_output:
visualization.enable_videowriter(
os.path.join(args.output_dir, "%s.avi" % args.sequence))
application.play_hypotheses(visualization)
if args.output_dir is not None:
pymotutils.motchallenge_io.write_hypotheses(
os.path.join(args.output_dir, "%s.txt" % args.sequence),
application.hypotheses)
def main():
"""Main program entry point."""
args = parse_args()
devkit = kitti.Devkit(args.kitti_dir)
data_source = devkit.create_data_source(
args.sequence, kitti.OBJECT_CLASSES_PEDESTRIANS,
min_confidence=args.min_confidence)
visualization = pymotutils.MonoVisualization(
update_ms=kitti.CAMERA_UPDATE_IN_MS,
window_shape=kitti.CAMERA_IMAGE_SHAPE)
application = pymotutils.Application(data_source)
# First, play detections. Then, show ground truth tracks.
application.play_detections(visualization)
application.play_track_set(data_source.ground_truth, visualization)
def main():
"""Main program entry point."""
args = parse_args()
devkit = motchallenge.Devkit(args.mot_dir)
data_source = devkit.create_data_source(args.sequence, args.min_confidence)
# Compute a suitable window shape.
image_shape = data_source.peek_image_shape()[::-1]
aspect_ratio = float(image_shape[0]) / image_shape[1]
window_shape = int(aspect_ratio * 600), 600
visualization = pymotutils.MonoVisualization(
update_ms=data_source.update_ms, window_shape=window_shape)
application = pymotutils.Application(data_source)
# First, play detections. Then, show ground truth tracks.
application.play_detections(visualization)
application.play_track_set(data_source.ground_truth, visualization)