search

tfzhou / ProtoSeg

CVPR2022 (Oral) - Rethinking Semantic Segmentation: A Prototype View
https://arxiv.org/abs/2203.15102
MIT License
341 stars 38 forks source link

Question about code #3

Open zqing10 opened 2 years ago

zqing10 commented 2 years ago

May I ask when will the core code be released? I am so interested in your works!!!!!!!!!!!

gasharper commented 2 years ago

same question.

Zxl19990529 commented 2 years ago

+one

gaowq2017 commented 2 years ago

+1 BTW the work same to 'few shot' leanning that used prototype, Thanks

wenguanwang commented 2 years ago

@gaowq2017 pls read the paper more carefully.

As we mentioned, our work shows the promise for formulating general segmentation and few-shot in a unified view of prototype learning. Ps: Prototype learning is a quite huge area (what a pity someone directly links it to few-shot learning).

Our work also shows an appealing direction -- the combination of supervised learning and unsupervised, within-class learning (clustering) leads to a more powerful pixel-classification framework.

We will have some more amazing results, let's wait and see.

gaowq2017 commented 2 years ago

@wenguanwang Thanks for your great work. I am agree that Prototype learning isn't only few-shot learning. I look forward to more similar work, thanks