search

Jyxarthur / flowsam

Official Implementation of "Moving Object Segmentation: All You Need Is SAM (and Flow)" Junyu Xie, Charig Yang, Weidi Xie, Andrew Zisserman
Apache License 2.0
219 stars 20 forks source link

More easy way to inference? #2

Open YiChenCityU opened 2 months ago

YiChenCityU commented 2 months ago

Could you provide more easy way to test your code? Such as just provide a mp4 as input? or a image directory?

Jyxarthur commented 2 months ago

Thanks for your interest in our work. I am afraid that currently the direct video input is not supported, and the alternative way is to convert the video file into images and store them into a directory. The detailed directory structure can be found in the "Inference" Section under the paragraph starting with "To run the code on your own data".

surya-narayanan commented 2 months ago

Thanks for the great tool. Two questions - 1. Is there a way to go from only the RGB to the output masks, and 2. Does this have the other powers of SAM, like specifying a point in the object I want to track, or specifying what I want to track using natural language?

Thanks again for a great tool!

SangbumChoi commented 2 months ago

https://github.com/Jyxarthur/flowsam/assets/34004152/ca59b1b7-bf62-417c-abd0-da2fa0184fe7

bird

@YiChenCityU You can try my open-PR