Thanks for your work! Given with a `pre-trained` model, how to perform single inference on image? Thanks,