search

hkchengrex / XMem

[ECCV 2022] XMem: Long-Term Video Object Segmentation with an Atkinson-Shiffrin Memory Model
https://hkchengrex.com/XMem/
MIT License
1.72k stars 191 forks source link

A simple question!Thanks #131

Closed 1334233852 closed 10 months ago

1334233852 commented 10 months ago

Thank you for your code and hard work！I have a simple question，batch_size = 8,The input data dimension accepted by KeyEncoder is BxCxHxW (before flattening, it was BxCxTxHxW), and I trained with two GPUs. Is BxCxTxHxW equal to [4,3,3384384]? Do I understand correctly? Thank you!

hkchengrex commented 10 months ago

In pre-training, yes (missed a comma?).

1334233852 commented 10 months ago

sorry sorry ，i missed a comma.... lol I remember that the main training seems to be the same, right?