How to generate the required flow vector for IDN from the flow stack extracted by TSN

[dejungle]

I tried to run the TVSeries test using the live stream images. For each chunk of images (N=6), I got a flow stack of size (10,1024, 1, 1), which is interlaced in x- and y- direction, using the TSN. However, the flow vector for each chunk in the given file 'tvseries_anet2016_test_data.pickle' is of size (1024, ). So, how can I generate the required flow vector for IDU from the flow stack extracted by TSN? Thanks a lot.

[lenny-rgb]

Hi @dejungle , Did you manage to figure this out?

Thanks cc: @hjeun