Speck 2b#
Speck 2b is a specific version of Speck, which integrates a 128 * 128 resolution DVS and the chip itself is fully configurable with the spiking neuron capacity of 32 million.
Key features#
1x Event (or DVS) Pre-Processing layer (incl. noise filtering)
9x Convolutional layers (incl. pooling)
1x Readout layer operating in average, max, max average, or threshold mode
Asynchronous serial interface
Ultra-low average working power consumption
DVS Layer#
128 x 128 Pixel Array
Sum pooling 1:1, 1:2, 1:4
1 bit conf per pixel (1 kill)
ROI selection
Noise filter and hot pixel rejection
On only / Off only / Both / Merge filtering of DVS event polarity
Mirroring in both X/Y
Rotate in 90-degree steps
Fanout of 2
CNN Layer#
Max input dimension 128*128
Max feature output size 64*64
Max feature number 1024
Weight resolution 8 bit
Neuron state resolution 16
Max kernel size 16*16
Stride {1,2,4,8} independent in X/Y
Padding [0..7] independent in X/Y
Pooling 1:1, 1:2, 1:4
Fanout of 2
Leak operation
Readout Layer#
16 classes
Selectable moving average between 1, 6 and 32 time steps.
4 readout modes: inactive/threshold/max spiking class/specific class.
Note
The readout layer inside the Speck 2b chip is not available at the moment
Api reference:#
Model: samna.speck2b
Event: samna.speck2b.event
Configuration: samna.speck2b.configuration