FiftyOne Recipes

Welcome to FiftyOne recipes!

FiftyOne turbocharges your current workflows, transforming hours of scripting into minutes so that you can focus on your models. Browse the recipes below to see how you can leverage FiftyOne to enhance key parts of your machine learning workflows.

Remove duplicate images from a dataset

Turn your data into a FiftyOne Dataset and automatically find and remove duplicate and near-duplicate images from your dataset.

# Find duplicates
dup_view = (
    # Extract samples with duplicate file hashes
    .match({"file_hash": {"$in": dup_filehashes}})
    # Sort by file hash so duplicates will be adjacent

# Visualize in App

Add model predictions to a datasets

Add FiftyOne to your model training and analysis loop to visualize and analyze your model’s predictions.

for img, sample_id in your_data:
    # Perform prediction
    label, confidence = your_model.predict(img)

    # Add prediction to FiftyOne dataset
    sample = dataset[sample_id]
    sample["your_model"] = fo.Classification(
        label=label, confidence=confidence,