generative-models
generative-models copied to clipboard
CLIP-Score Calculation
Hi, thank you for the amazing work of SV4D. I have a question about if the reported metric CLIP-s is calculated by:
- The generated text and image pair
- The similarity of GT image and generated image like Consistent4D
Additionally, if it is the first one, how do you decide which text is reliable? Thank you!
Sincerely, Chih-Chuan