Add option to avoid frame slicing of the feature representation
Closed this issue · 1 comments
pzinemanas commented
In some contexts would be interesting not to slice the feature representation into frames. Although this process was devised for fix-length-input neural network, in other contexts can be unnecessary or problematic. One possible solution is to avoid slicing the feature representation if sequence_time<=0
.
pzinemanas commented
Closed in 5786ab6