hkchengrex / Mask-Propagation

[CVPR 2021] MiVOS - Mask Propagation module. Reproduced STM (and better) with training code :star2:. Semi-supervised video object segmentation evaluation.
https://hkchengrex.github.io/MiVOS/
MIT License
127 stars 22 forks source link

How do I get the penultimate frame? #30

Closed longmalongma closed 3 years ago

longmalongma commented 3 years ago

image

How do I get the penultimate frame? Am I doing it right?

hkchengrex commented 3 years ago

I don't know... What are you trying to do? There should be three frames so Fs[:, 2] is the last one.

longmalongma commented 3 years ago

I don't know... What are you trying to do? There should be three frames so Fs[:, 2] is the last one.

I don't know... What are you trying to do? There should be three frames so Fs[:, 2] is the last one. Fs[:, 0]是当前帧,Fs[:, 1]是前一帧,Fs[:, 2]是前一帧的前一帧,这样的理解对吗?我给你发了邮件。

hkchengrex commented 3 years ago

No. Fs[:, 0] is the first frame, Fs[:, 1] is the second frame, and Fs[:, 2] is the third frame. They are temporally ordered but not consecutive. You can use tensorboard to visualize the output.

You need to modify the datasets in order to load consecutive frames (see vos_dataset.py).

longmalongma commented 3 years ago

No. Fs[:, 0] is the first frame, Fs[:, 1] is the second frame, and Fs[:, 2] is the third frame. They are temporally ordered but not consecutive. You can use tensorboard to visualize the output.

You need to modify the datasets in order to load consecutive frames (see vos_dataset.py).

Thanks, in that case,which one is the current frame?

hkchengrex commented 3 years ago

We perform two steps here. In the first step Fs[:, 1] is the current frame; in the second step Fs[:, 2] is the current frame.

longmalongma commented 3 years ago

We perform two steps here. In the first step Fs[:, 1] is the current frame; in the second step Fs[:, 2] is the current frame.

We perform two steps here. In the first step Fs[:, 1] is the current frame; in the second step Fs[:, 2] is the current frame.

Ok, I got it, thank you very much!