Closed YY-Wei closed 5 months ago
gpt4v.py takes 3 images generated by Merge
, Switch
, and Composite
respectively as input. It then performs 4 comparative evaluations (Merge
vs. Switch
, Switch
vs. Merge
, Merge
vs. Composite
, and Composite
vs. Merge
) to calculate a score for each image.
For explanations of evaluation, you can refer to evaluate.py and eval.sh.
It seems that Explanation of evaluation in gpt4v.py cannot work, gpt4v will not output why.