This post shows how to load and evaluate the model we built in the previous post.

import dill
from siam_utils import SiameseImage
from sklearn.metrics import accuracy_score

from fastai.vision.all import *

learner = load_learner(Path(os.getenv("MODELS")) / "siam_catsvdogs_tutorial.pkl", cpu=False, pickle_module=dill)

path = untar_data(URLs.PETS)
files = get_image_files(path / "images")

imgval = PILImage.create(files[0])
imgtest = PILImage.create(files[1])
siamtest = SiameseImage(imgval, imgtest)

siamtest.show();

@patch
def siampredict(self: Learner, item, rm_type_tfms=None, with_input=False):
    res = self.predict(item, rm_type_tfms=None, with_input=False)
    if res[0] == tensor(0):
        SiameseImage(item[0], item[1], "Prediction: Not similar").show()
    else:
        SiameseImage(item[0], item[1], "Prediction: Similar").show()
    return res

res = learner.siampredict(siamtest)

imgval = PILImage.create(files[9])
imgtest = PILImage.create(files[35])
siamtest = SiameseImage(imgval, imgtest)

siamtest.show();

res = learner.siampredict(siamtest)

Let’s see how well it does overall.

test_dl = learner.dls.test_dl(files)

preds, ground_truth = learner.get_preds(dl=test_dl)

round(accuracy_score(ground_truth, torch.argmax(preds, dim=1)), 4)

0.9306