seems incorrect result

by xudou003 - opened 9 days ago

9 days ago

i try the demo code, it seems correct for the demo image which is a diagram. But I change to a dog picture, it still outputs diagram with highest confidence. Then I test MobileCLIP2-S4 with same code, it works. For MobileCLIP2-S2 I meet the same problem, any suggestion?
here is my picture, it will predict it as "a diagram".

fartashf

Apple org 7 days ago

Hi @xudou003 ,
Can you confirm if you have used the correct normalization (mean=0, std=1)? The normalization for S3/S4/L-14 variants are different from S0/S2/B variants. See #2 and #3 and our instructions in the GitHub Readme. Please also make sure that the model is in eval mode.

xudou003

7 days ago

Thank you!

xudou003 changed discussion status to closed 7 days ago

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment