tracking

TrackingData

class stardust.metric.tracking.TrackingData(track_id, box_type, obj_info)[source]

Bases: object

Define a tracking object

Args:

track_id: int: tracking id of object
box_type: str: type of object, which should be chosen from ‘2D’ and ‘3D’
obj_info: List: info of box, [x, y, z, l, w, h, ry] for box 3d and [x, y, w, h] for box 2d

Returns:

TrackingData object

Examples:

from stardust.metric.tracking import TrackingData
t_data = TrackingData(0, '3D', [1, 2, 3, 4.5, 1.8, 1.5, 0.1])
t_data = TrackingData(0, '2D', [1, 2, 4.5, 1.8])

TrackingMetric

class stardust.metric.tracking.TrackingMetric(trajectory_gt, trajectory_pd, IoU_thr, metric_tyoe)[source]

Bases: object

Define a TrackMetric object and prepare metric data

Args:

trajectory_gt: List: trajectory of annotation
trajectory_pd: List: trajectory of prediction
IoU_thr: float: The iou threshold of tp boxes
metric_tyoe: str: type of input object, which should be chosen from ‘2D’ and ‘3D’

Returns:

TrackingMetric obj

compute()[source]

compute tracking metric

Args:

Returns:: Dict

compute_metric

stardust.metric.tracking.compute_metric(data=None, IoU_thr=None, metric_types=None, save_path=None)[source]

Computing IoU of all objects in all frames

Args:

data: Generator: A generator object to get all information from all frames
IoU_thr: float: The iou threshold of tp boxes
metric_types: List: which type of box to compute metric, which can be chosen from [‘2D’], [‘3D’] and [‘2D’, ‘3D’]
save_path: str: Local path to save metric results

Returns:

metric: List: The metric of dataset which include two dict, the first represents metric of every single frame and the second represents metric of all frames

Examples:

from stardust.metric.tracking import compute_metric
project_id = 856
Export(project_id, 'top', input_path, True).export()
json_datas = read_rosetta(project_id=project_id,
                        input_path=input_path,
                        )
metric = compute_metric(json_datas, 0.5, ['2D', '3D'], 'local/')