search

mit-han-lab / fastcomposer

[IJCV] FastComposer: Tuning-Free Multi-Subject Image Generation with Localized Attention
https://fastcomposer.mit.edu
MIT License
658 stars 37 forks source link

Appendix + Evaluation Configs #11

Closed darkliang closed 1 year ago

darkliang commented 1 year ago

Hello, I saw that the paper claims that some details are in the appendix, but I can't find the appendix, can you provide it?

tianweiy commented 1 year ago

@darkliang you can find it here appendix.pdf

Evaluation details from the appendix: In our experiments, we use α = 0.5 for single object generation, and α = 0.6 for multi-subject generation. We use PNDM sampling [3] with 50 steps and a classifier-free guidance scale of 5 across all methods.

we will put it online soon. Thanks

darkliang commented 1 year ago

Thanks a lot!

darkliang commented 1 year ago

Hi, I read your appendix, but I couldn't find all the images you used in your evaluation.

tianweiy commented 1 year ago

we plan to release the data + evaluation code but it is going to take some time to prepare. stay tuned