recognize-anything
recognize-anything copied to clipboard
Inference_ram.py example using Tag2Text
Hello,
I am confused. It seems ithe nference_ram.py example is using
default='pretrained/tag2text_swin_14m.pth'
which is the same as the model used in the inferecen_tag2text.py.
Can someone clarify if this was not a typo?
Sorry and thanks for pointing out, yes, this is a typo. It should be pretrained/ram_swin_large_14m.pth
.
Please follow the command in README, i.e.,
python inference_ram_openset.py \
--image images/openset_example.jpg \
--pretrained pretrained/ram_swin_large_14m.pth
We will correct it soon.