You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi,
I tried classify_oidv2.py on many images and it works great. But I noticed that some classes are not predicted even it's very clear from the image. The classes are among trainable classes. E.g. I tried several images of "Perfume" and "Rafting" and "Perfume"/"Rafting" was not among predicted labels. Otherwise the model works great. Any idea why these labels are not predicted? Maybe these labels was not so frequent in the training data? Thanks for any comment or insight.
The text was updated successfully, but these errors were encountered:
I tried prediction on 25,000 testing images (having at least 10 images per trainable class) and I got only about 1,800 classes. That means about 3,200 classes weren't predicted even that they are trainable. Please, any idea why it happens? @nalldrin
Hi,
I tried classify_oidv2.py on many images and it works great. But I noticed that some classes are not predicted even it's very clear from the image. The classes are among trainable classes. E.g. I tried several images of "Perfume" and "Rafting" and "Perfume"/"Rafting" was not among predicted labels. Otherwise the model works great. Any idea why these labels are not predicted? Maybe these labels was not so frequent in the training data? Thanks for any comment or insight.
The text was updated successfully, but these errors were encountered: