tracklab.wrappers.eval package

Subpackages

• tracklab.wrappers.eval.posetrack package
  • Submodules
  • tracklab.wrappers.eval.posetrack.posetrack18_evaluator module
    • PoseTrack18Evaluator
      • PoseTrack18Evaluator.run()
  • tracklab.wrappers.eval.posetrack.posetrack21_evaluator module
    • PoseTrack21Evaluator
      • PoseTrack21Evaluator.PoseTrackEncoder
      • PoseTrack21Evaluator.check_if_tracklet()
      • PoseTrack21Evaluator.compute_bbox_map()
      • PoseTrack21Evaluator.format_metric()
      • PoseTrack21Evaluator.has_enough_vis_kp()
      • PoseTrack21Evaluator.remove_not_visible_kps()
      • PoseTrack21Evaluator.run()
• tracklab.wrappers.eval.soccernet package
  • Submodules
  • tracklab.wrappers.eval.soccernet.soccer_accuracy module
    • SoccerAccuracy
      • SoccerAccuracy.run()

Submodules

tracklab.wrappers.eval.trackeval_evaluator module

class tracklab.wrappers.eval.trackeval_evaluator.TrackEvalEvaluator(cfg, eval_set, show_progressbar, dataset_path, tracking_dataset, *args, **kwargs)

Bases: Evaluator

Evaluator using the TrackEval library (https://github.com/JonathonLuiten/TrackEval). Save on disk the tracking predictions and ground truth in MOT Challenge format and run the evaluation by calling TrackEval.

run(tracker_state)

Run the evaluation :param tracker_state: the tracker state for the evaluation :type tracker_state: TrackerState

tracklab.wrappers.eval.trackeval_evaluator.format_metric(metric_name, metric_value, scale_factor)