Sequence prediction

[abduallahmohamed]

Sequence prediction

[abduallahmohamed]

For GTA-IM we used the 5 FPS version following prior works. I don't recall the FPS for the other dataset but we used the default version.
For observation we followed prior works of observing 5 sequences then the prediction can start with at least 10 sequences based on how long you want to predict into the future. This is why you see our evaluation metrics account for different prediction horizons.

[MZ82020]

How is this procedure done? I don't understand much.. We observe 5 sequences and predict at least 10 sequences, and the FPS5 folder contains 14 sequences. How is this?

I don't have much how you did it and where it is modeled what you said in the code!

[abduallahmohamed]

The Sequences in the GTA-IM folder are already sampled. Please check the dataloader:
https://github.com/abduallahmohamed/Skeleton-Graph/blob/main/gtaimdset.py
It creates a sliding window

Skeleton-Graph/gtaimdset.py

Line 150 in c6f37ec

#Create the sequences, using a moving window of (seq_in_seq_out)

[MZ82020]

x, xA, y represent what please?

[abduallahmohamed]

x is the 2d input joins xA is the adjacency matrix y is the 3d target poses ------------------------------------------ Abduallah Mohamed abduallahmohamed.com

…

On Mon, 28 Nov 2022 at 00:57, MZ82020 ***@***.***> wrote: x, xA, y represent what please? — Reply to this email directly, view it on GitHub <#4 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFIVLG7ZWECUMNPSHWMQDMLWKRXXZANCNFSM6AAAAAASA74LWA> . You are receiving this because you commented.Message ID: ***@***.***>