xai4tsc.utils.animation

Perturbation-process animations (GIF) for the frequency / time-frequency metrics.

Each animation shows a single sample’s perturbation trajectory over the metric’s flipping steps: the reconstructed time signal, the perturbed frequency / time-frequency representation (with the static relevance), and the prediction curve building up. Frames are produced by the perturbation metrics’ animate method (opt-in frame collection); output is always a GIF via Matplotlib’s built-in PillowWriter (no ffmpeg dependency).

Attributes

logger

Classes

FrequencyPerturbationAnimation	Animate a frequency-domain perturbation: spectrum + time signal + prediction.
TimeFrequencyPerturbationAnimation	Animate a TF perturbation: relevance + spectrogram + time signal + prediction.

Module Contents

xai4tsc.utils.animation.logger

class xai4tsc.utils.animation.FrequencyPerturbationAnimation(frames: dict, step: int = 1, max_frames: int = 300)

Bases: _PerturbationAnimation

Animate a frequency-domain perturbation: spectrum + time signal + prediction.

class xai4tsc.utils.animation.TimeFrequencyPerturbationAnimation(frames: dict, step: int = 1, max_frames: int = 300)

Bases: _PerturbationAnimation

Animate a TF perturbation: relevance + spectrogram + time signal + prediction.