You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently it is not straightforward to record data from hardware (either from the vision sensor or layer outputs) without providing any input.
For example, if I just want to record data from the speck sensor for a given duration, I would not know how to do this with sinabs-dynapcnn (without writing lower-level samna code).
The text was updated successfully, but these errors were encountered:
There is a way that is probably not documented, nor tested:
Instantiate a DynapcnnNetwork with a sequential that only contains a DVSLayer.
Deploy that network onto chip with the to method, passing monitor_layers = ["dvs"]
To record, call the forward method of the DynapcnnNetwork instance. As argument pass a list with one single dummy event that has as timestamp the desired recording duration (in microseconds).
The return value will be a list of all events recorded from the DVS within the specified duration.
This issue could probably be solved by adding proper documentation to the procedure above and a unit test that ensures it works.
Currently it is not straightforward to record data from hardware (either from the vision sensor or layer outputs) without providing any input.
For example, if I just want to record data from the speck sensor for a given duration, I would not know how to do this with sinabs-dynapcnn (without writing lower-level samna code).
The text was updated successfully, but these errors were encountered: