Skip to content

Example Classify Order#

To learn about the basics of creating an order, please refer to the quickstart guide.

Order description#

With this order we want to find out what kind of emotions certain AI generated image convey. We have asked Dalle-3 to generate 4 different images that would convey happiness, anger, disgust and sadness and saved those images to publically accessible URLs.

Now we want to find out if the images actually match the emotions we asked for. When you run this with your own examples, you may use local paths to your images instead of the URLs.

'''
Classify order with a validation set
'''

from rapidata import RapidataClient

def get_urls():
    base_url = "https://assets.rapidata.ai/dalle-3_"
    emotions = ["anger", "disgust", "happiness", "sadness"]
    generated_images_urls = [f"{base_url}{emotion}.webp" for emotion in emotions]
    return generated_images_urls


def get_emotions_of_images_order(rapi: RapidataClient):
    generated_images_urls = get_urls()

    # Configure order
    order = rapi.order.create_classification_order(
        name="emotions from images",
        instruction="What emotions do you feel when looking at the image?",
        answer_options=["happy", "sad", "angry", "surprised", "disgusted", "scared", "neutral"],
        datapoints=generated_images_urls,
        responses_per_datapoint=50
        ).run()

    return order


if __name__ == "__main__":
    order = get_emotions_of_images_order(RapidataClient())
    order.display_progress_bar()
    results = order.get_results()
    print(results)

The resulting rapids for the users look like this:

Classify Example