Passing Event Camera Event-Train

[danielmimimi]

First - the implementation works good, but I have a lack of understanding concerning the asynchronous input.

If I have an event train of position X,Y at ts (I ignore the Polarity for now), how would I pass the stream into the PhasedLSTM?

In the paper : https://arxiv.org/abs/1610.09513 they used kind of an embedding with 40 dimensions (from which the source is not really helpful) and pass the EventMnist into the PLSTM.

In your example you pass the MNIST image as one dimensional input PLSTM(32, input_shape=(28 * 28, 1), implementation=2) which is ok but I think its not the way the PLSTM should be used. Because what frequency will be learned ?

Can anybody help me?