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
Can be useful for dense multichannel true/false labels. One can just use UInt8's, but if only true/false is allowed then Bool is a natural representation for it, and moreover if whatever process is generating these labels outputs Matrix{Bool}, then if Onda supported Bool, it could be wrapped in Samples and encoded/decoded without copying.
Onda allows us to skip the copy if sample_type === eltype(sample_data) && sample_resolution_in_unit == 1 && sample_offset_in_unit == 0 but that means the label generating process would need to output UInt8's instead of Bools which can be less natural.
Additionally, maybe Onda could bitpack the Bools i.e. support a storage format like BitMatrix for 8x storage savings, although that sounds like a bigger lift to me.
The text was updated successfully, but these errors were encountered:
Can be useful for dense multichannel true/false labels. One can just use
UInt8
's, but if only true/false is allowed thenBool
is a natural representation for it, and moreover if whatever process is generating these labels outputsMatrix{Bool}
, then if Onda supportedBool
, it could be wrapped inSamples
and encoded/decoded without copying.Onda allows us to skip the copy if
sample_type === eltype(sample_data) && sample_resolution_in_unit == 1 && sample_offset_in_unit == 0
but that means the label generating process would need to outputUInt8
's instead ofBool
s which can be less natural.Additionally, maybe Onda could bitpack the
Bool
s i.e. support a storage format likeBitMatrix
for 8x storage savings, although that sounds like a bigger lift to me.The text was updated successfully, but these errors were encountered: