Tracked Persons
0
active tracks
Spikes / sec
0
Energy Saved
87 %
Active Tracks
0
vs dense baseline
Membrane τ
8 ms
leak constant
Neurons
2,048
LIF · temporal code
⚡ Cumulative Energy Saved
SNN sparsedense baseline
🏃 Energy Efficiency
Spike sparsity3.1%
Idle neurons96.9%
Codingtemporal (TTFS)
🧩 Energy Budget
👤 Tracked Persons
live IDs🔌 System Status
NetworkSpiking (LIF)
CSI linkStable
Refractory2 ms
Thresholdv_th 1.0
Modelspike-track v0.5
🔭 Spike Raster
128 neuronssparse membrane firings · pink = spike this window
⚡ Energy Saved
Avg spikes/neuron0.03
Inference clockgated
📈 Spikes per Second
live👤 Tracked Persons
0 tracks🌡 Membrane Potential Heatmap
layer × timerows: SNN layers L1–L6 · cols: time → now
Avg Energy Saved
84 %
vs dense CNN
Peak Tracks
6
concurrent
Mean Spikes/s
142
network-wide
Track Continuity
93 %
re-ID success
🔌 Energy Saved per Hour
24h📐 Spike-rate Distribution
🌡 Activity by Hour × Day
7×24🧩 Compute Split
Sparse spikes11%
Gated idle84%
Routing/overhead5%
📋 Event Log
All
| Time | Event | Severity | Detail |
|---|
🔔 Live Stream
tail🎛 SNN Tuning
8 ms
Larger τ integrates over longer windows; smaller τ reacts to fast motion spikes.
1.0
2 ms
⚡ Power & Tracking
Event-driven clock gating
Halt neurons with no input spikes
Persistent person re-ID
Maintain track IDs across gaps
Deep sleep when idle
Sleep SNN when no motion 30 s
Publish energy_saved events
Report savings to Seed bus
📋 Cog Details
DescriptionImplements spiking neural network for energy-efficient person tracking using temporal spike coding of CSI events.
Events
HardwareESP32-S3
InputCSI motion
Versionv0.5.0 · 26 KB · Hard