Configuration

TemporalFixConfig is a frozen validated dataclass. Construct it directly, load strict YAML with load_config, or resolve an inspectable preset.

from temporalfix import BoxSmoothing, TemporalFixConfig

config = TemporalFixConfig(
    max_missing_frames=3,
    min_iou=0.35,
    box_smoothing=BoxSmoothing.EMA,
    output_tentative=True,
)
print(config.to_dict())

Available presets are balanced, low_latency, high_stability, and strict_false_positive_control. No values are hidden:

temporalfix inspect-config --preset high_stability --format yaml

Fields

Area	Fields
Association	min_iou, class_gating
Gap lifecycle	max_missing_frames
Box smoothing	box_smoothing, ema_alpha, kalman_process_noise, kalman_measurement_noise
Confidence	confidence_stabilization, confidence_alpha, confidence_decay, min_confidence
Class evidence	class_stabilization, class_voting, class_history_size, class_evidence_decay, class_switch_threshold
Confirmation	suppress_short_tracks, min_confirmed_observations, max_tentative_age, output_tentative
Uncertainty	initial_uncertainty, uncertainty_growth

YAML is limited to 1 MiB, parsed with yaml.safe_load, and must contain a mapping. Unknown fields and invalid ranges raise clear errors.