question about test.pkl/train.pkl file

[qdinfish]

Hi,

I'm reading your paper, thanks for uploading your code for study.

I have a quick question, as you mentioned in readme. The .pkl need have format as

120 rows (12 person *10 steps) , and each row has 4096 .

And following 2 function could hanlde the .pkl file , no need code change, right?

semantic_ae.network_build_layers_v1() or semantic_ae.network_build_layers_v2()

[mostafa-saad]

The provided code could be run and minimum so that you could trace.

[qdinfish]

So I need do some changes to meet the 10 steps temporal case ? I mean 120 rows (12 person *10 steps).

[mostafa-saad]

yes, re-read the README page for the notes.

[qdinfish]

OK, thanks a lot. I see "window=1" is set in code. For the data format, the target shall be the label of the activity, right?

[mostafa-saad]

yes

[qdinfish]

How can I extract the person‘s feature after relation layer？ eg. save them to a file

[mostafa-saad]

The layer output is the K players features directly. Put effort to understand the code.