search

hkchengrex / MiVOS

[CVPR 2021] Modular Interactive Video Object Segmentation: Interaction-to-Mask, Propagation and Difference-Aware Fusion. Semi-supervised VOS as well!
https://hkchengrex.com/MiVOS/
MIT License
469 stars 64 forks source link

Is that possible this work in real-time case #17

Closed mathmax12 closed 3 years ago

mathmax12 commented 3 years ago

Hi Thanks for sharing the amazing work. Does the work support real-time case? If not what change do you recommend for that? Thanks.

hkchengrex commented 3 years ago

Can you elaborate on what properties do your real-time case needs? (speed/online/long sequences?)

mathmax12 commented 3 years ago

For example, processing the video stream from a camera in real-time 30FPS.

hkchengrex commented 3 years ago

Speed-wise, the current model can handle it with PyTorch's AMP enabled in the single object setting. But then there are also considerations regarding inputs (this is after all an interactive method) and memory (feature bank growing over time as in STM).

mathmax12 commented 3 years ago

I see. Thanks for that.

On Wed, Sep 8, 2021 at 6:46 PM Rex Cheng @.***> wrote:

Speed-wise, the current model can handle it with PyTorch's AMP https://pytorch.org/docs/stable/amp.html enabled in the single object setting. But then there are also considerations regarding inputs (this is after all an interactive method) and memory (feature bank growing over time as in STM).

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/hkchengrex/MiVOS/issues/17#issuecomment-915693990, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHW6H62U7WPB5ORSP5PTYQLUBAGVPANCNFSM5DVUHWZQ . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.