Can this method perform in-context image segmentation?

edward3862 / Analogist

Analogist: Out-of-the-box Visual In-Context Learning with Image Diffusion Model (SIGGRAPH 2024)

https://analogist2d.github.io

MIT License

33 stars 1 forks source link

Can this method perform in-context image segmentation? #1

Open jianzongwu opened 2 months ago

jianzongwu commented 2 months ago

Hi @edward3862

Thanks for your great work!

I want to know, can this method perform in-context image segmentation?

edward3862 commented 2 months ago

Hi @jianzongwu

Thank you for your attention to our work!

If it is necessary to perform segmentation, our method will produce a mask. Note that the Diffusion Model we use is primarily pre-trained for generating natural images; therefore, the accuracy cannot be fully guaranteed.