This (mostly) adds a function models.wrappers.create_image_feature_extractor that will return model object that "predicts the image features" instead of the classes (see the documentation). You can either pass in a BigML image deepnet or the name of a pretrained model (e.g., "resnet18").
This (mostly) adds a function
models.wrappers.create_image_feature_extractor
that will return model object that "predicts the image features" instead of the classes (see the documentation). You can either pass in a BigML image deepnet or the name of a pretrained model (e.g., "resnet18")./cc @unmonoqueteclea @aclementev