search

rdroste / unisal

Unified Image and Video Saliency Modeling (ECCV 2020)
https://arxiv.org/abs/2003.05477
Apache License 2.0
128 stars 36 forks source link

enabled CPU inference #19

Closed ohjho closed 2 months ago

ohjho commented 2 months ago

changed the code in a few places so that it's able to run on CPU

also changed the code formatting a little using Black

rdroste commented 2 months ago

Thx for the contribution!

ohjho commented 2 months ago

and thanks for the great model!

i'm using it for a project and making a few additions to make inference of images and videos more streamlined, if you would like I could push those changes into here too

rdroste commented 1 month ago

Thanks for the suggestion. Feel free to open a PR once you have those changes. However, note that this repository needs to replicate the results from our paper, so I can't include changes that affect the accuracy. If the changes are too far-reaching, I can always link to your fork in the README.

ohjho commented 1 month ago

yes totally understand, the latest PR #21 only added inference functions, no subtractions, no modification of any of your existing methods or functions (aside from the code formatting for "better" readability).

I focused mostly on image inference for now, will make similar inference support for videos in a future PR

i'm currently testing the use of your model in combination of several different foundation models (e.g. SAM) to get salient objects