search

AILab-CVC / SEED

Official implementation of SEED-LLaMA (ICLR 2024).
https://ailab-cvc.github.io/seed
Other
515 stars 30 forks source link

CIDEr reproduce hyperparameter #9

Closed zheedong closed 7 months ago

zheedong commented 7 months ago

Hello, thank you for your research and I'm very interested in it. I'd like to reproduce your research results, so could you tell me about the specific evaluation setting?

geyuying commented 7 months ago

Thank you for your interest in our work. For evaluating image captioning, we use model.generate(num_beams=5, do_sample=False).

zheedong commented 7 months ago

Hi, thanks for your reply. Thanks to your reply, it was successful to reproduce your CIDEr results. I have some other questions about your works.

  1. Which hyperparameter do you use for CLIP Score image generation?
  2. Do you have any measurement about COCO test set FID?

Thank you.