Skip to content

Loading image classification model for inference

geekjr edited this page Apr 17, 2021 · 2 revisions

In the previous tutorial, we trained a network that could identify between 2 different kids of cars. Here, we will use that model for inference.

Start by downloading a picture that is of one of the 2 cars. Here is an example.

First create a new Python script called image_classification_cars_prediction.py. Make sure the image you downloaded is in the same folder as this new file. Open up that file and start by importing the ImageClassificationPredictor class from quickai:

from quickai import ImageClassificationPredictor

Once you have imported that class, lets call it to make our prediction:

predictions = ImageClassificationPredictor(
    "cars", 224, "00198.jpg", [
        'Acura Integra Type R 2001', 'Acura RL Sedan 2012'])
print(predictions)

In the above line of code we call the ImageClassificationPredictor with cars as the save, 224 as dims, 00198.jpg as path, and [ 'Acura Integra Type R 2001', 'Acura RL Sedan 2012'] as classes. save is the parameter that specifies the name of the model to load, dims is the dimensions the model needs for prediction(224 in the case of VGG16), path is the path of the image to use for prediction, and classes is a list of our classes.

Now, go ahead and run image_classification_cars_prediction.py. After a few seconds you will see the predictions printed to the console!