search

hkchengrex / Tracking-Anything-with-DEVA

[ICCV 2023] Tracking Anything with Decoupled Video Segmentation
https://hkchengrex.com/Tracking-Anything-with-DEVA/
Other
1.27k stars 129 forks source link

Didn't you use a mask that was inferred from a large model during semi-youtubevos inference? #75

Closed longmalongma closed 7 months ago

longmalongma commented 7 months ago

Didn't you use a mask that was inferred from a large model during semi-youtubevos inference? I see the script is not loading

image

hkchengrex commented 7 months ago

No. This is for semi-supervised VOS. Only the first-mask input is used.

longmalongma commented 7 months ago

No. This is for semi-supervised VOS. Only the first-mask input is used.

Okay, for semi vos, how do we use the mask inferred by Sam? Is there a script like this?@hkchengrex

hkchengrex commented 7 months ago

We didn't test this application.