efficient way to customize the model for more types of images

[ursvasan]

I have a use case where I have some type of images where this model doesn't work well, but works well for many of my images. If I want to add some more images to the training data and retrain what is the most efficient approach ?