Visualization Guide

SpikeSEG includes 18 matplotlib-based plot functions in spikeseg.utils.visualization. All accept an optional save_path to write a PNG/PDF file.

Available Plots

Function	Description
plot_learned_filters	Grid of learned STDP convolutional filters
plot_filter_evolution	How filters change across training epochs
plot_saliency_map	Decoder output heat map
plot_saliency_overlay	Saliency overlaid on input image
plot_segmentation_result	Prediction vs ground truth side by side
plot_spike_raster	Neuron spike times over time
plot_membrane_traces	Membrane potential traces with threshold line
plot_spike_activity_map	Spatial spike counts per feature map
plot_events_frame	Single accumulated event frame
plot_time_surface	Time-surface representation of events
plot_event_sequence	Sequence of event frames across timesteps
plot_convergence_metric	$C_l$ over training steps
plot_weight_distribution	Histogram of synapse weights
plot_wta_wins	Bar chart of WTA wins per feature
plot_training_metrics	Multi-panel training dashboard
plot_feature_activations	Grid of feature map activations
plot_satellite_detection	Detections overlaid on event frame

Style Presets

from spikeseg.utils.visualization import set_style

set_style("paper")   # Clean, publication-ready defaults

Examples

Learned Filters

from spikeseg.utils.visualization import plot_learned_filters

weights = model.encoder.conv2.weight.detach().cpu()
plot_learned_filters(weights, title="Conv2 STDP Filters", save_path="filters.png")

Saliency Map

from spikeseg.utils.visualization import plot_saliency_map

plot_saliency_map(saliency[0, 0].cpu(), save_path="saliency.png")

Convergence

from spikeseg.utils.visualization import plot_convergence_metric

plot_convergence_metric(
    learner.convergence_history,
    threshold=0.01,
    save_path="convergence.png",
)

Spike Raster

from spikeseg.utils.visualization import plot_spike_raster

# spikes: (T, N) or (T, C, H, W) -- automatically flattened
plot_spike_raster(spikes, title="Conv2 Spike Raster", save_path="raster.png")

Configuration

VisualizationConfig controls defaults (figsize, DPI, colormaps, font sizes, FPS):

from spikeseg.utils.visualization import VisualizationConfig

cfg = VisualizationConfig(figsize=(16, 10), dpi=200, cmap_saliency="inferno")

TensorBoard Integration

Plots can be logged to TensorBoard via TensorBoardLogger.log_figure:

from spikeseg.utils.logging import TensorBoardLogger

tb = TensorBoardLogger("runs/experiment/events")
fig = plot_learned_filters(weights)
tb.log_figure("filters/conv2", fig, step=epoch)