Skip to content

Humans Problem


You've developed a 🤖 machine learning model that classifies images. Specifically, it outputs labels with non-negligible probabilities.

import pandas as pd

predictions = pd.DataFrame.from_dict({
    'preds': {
        12: "{'dog': 0.55, 'cat': 0.25, 'squirrel': 0.2}",
        41: "{'telephone pole': 0.8, 'tower': 0.1, 'stick': 0.1}",
        43: "{'man': 0.65, 'woman': 0.33, 'monkey': 0.02}",
        46: "{'waiter': 0.45, 'waitress': 0.30, 'newspaper': 0.15, 'cat': 0.10}",
        49: "{'nurse': 0.50, 'doctor': 0.50}",
        72: "{'baseball': 0.8, 'basketball': 0.15, 'football': 0.05}",
        91: "{'woman': 0.62, 'man': 0.28, 'elephant': 0.10}"
    }
})

print(predictions)
#                                                                  preds
# 12                         {'dog': 0.55, 'cat': 0.25, 'squirrel': 0.2}
# 41                 {'telephone pole': 0.8, 'tower': 0.1, 'stick': 0.1}
# 43                        {'man': 0.65, 'woman': 0.33, 'monkey': 0.02}
# 46  {'waiter': 0.45, 'waitress': 0.30, 'newspaper': 0.15, 'cat': 0.10}
# 49                                     {'nurse': 0.50, 'doctor': 0.50}
# 72             {'baseball': 0.8, 'basketball': 0.15, 'football': 0.05}
# 91                      {'woman': 0.62, 'man': 0.28, 'elephant': 0.10}

Each row in predictions represents predictions for a different image.

Insert a column called prob_human that calculates the probability each image represents a human. You can use the following list of strings to identify human labels.

humans = ['doctor', 'man', 'nurse', 'teacher', 'waiter', 'waitress', 'woman']
Prevent pandas from truncating print(predictions)

When you print(predictions), the output might get truncated like this.

To prevent this, set display.max_colwidth to None.

pd.set_option('display.max_colwidth', None)
print(predictions)


Try with Google Colab See the solution