Instructions to use Bedovyy/OppaiOracle-V1.1-timm with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- timm
How to use Bedovyy/OppaiOracle-V1.1-timm with timm:
import timm model = timm.create_model("hf_hub:Bedovyy/OppaiOracle-V1.1-timm", pretrained=True) - Notebooks
- Google Colab
- Kaggle
RGB to BGR
Did you tune it a little on the new color format? Accuracy takes a bit from what I understand unless that's tuned a little. I didn't know BGR was standard. Ive noted that for my next version. I'm currently in the process of building and cleaning my next dataset. I'll be training again once the months get cooler.
Thank you for the good model.
I just switched the first layer. (I uploaded the script)
BGR is not standard actually, it was because timm version of SmilingWolf's models still? use BGR and my toy project starts from it, so I switched color space of your model, instead switching SmilingWolf's.
Recent models seem to use RGB, so do not switch to BGR, because of this, please.
I am thinking that I keep color space of the model during convert, then add note on config.json so the tool can handle both RGB and BGR. (I already did it for padding)