AgriFieldNet India Challenge

Can you detect crop types in a class-imbalanced satellite image dataset?

Help on competition's metric

Hello! This competition is very interesting and while trying to evaluate my results in a validation set I struggled a little bit with the metric.

The Evaluation says:

"The evaluation metric for this challenge is Cross Entropy with the binary outcome for each crop"

I implemented a function but I'm not sure if this is the correct implementation. Could you please check if this is correct?

import pandas as pd

import numpy as np

def cross_entropy(y_true: pd.Series, y_prob: np.ndarray):

y_true = pd.get_dummies(y_true).values

# log is undefined for p=0 or p=1 y_prob = np.clip(y_prob, 1e-6, 1 - 1e-6)

return -(y_true * np.log(y_prob)).sum() / len(y_true)

If this is not the case, should I just use sklearn like this?

from sklearn.metrics import log_loss

log_loss(y_true, y_prob)

Discussion 0 answers