Is it unfair to use a pre-trained CLIP model compared to some other methods in Table 1? #24

maoSuxi · 2023-10-13T01:31:46Z

sry, a little problem, most baseline methods in table 1 use RN50 or ViT as backbone,
i think it's better to report the performance on RN50 or ViT to avoid the extra benefit that pre-trained large models brings？

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Is it unfair to use a pre-trained CLIP model compared to some other methods in Table 1? #24

Is it unfair to use a pre-trained CLIP model compared to some other methods in Table 1? #24

maoSuxi commented Oct 13, 2023

Is it unfair to use a pre-trained CLIP model compared to some other methods in Table 1? #24

Is it unfair to use a pre-trained CLIP model compared to some other methods in Table 1? #24

Comments

maoSuxi commented Oct 13, 2023