entropy module
- class datumaro.plugins.sampler.algorithm.entropy.SampleEntropy(data, inference)[source]
Entropy is a class that inherits an Sampler, calculates an uncertainty score based on an entropy, and get samples based on that score.
- __init__(data, inference)[source]
Constructor function
- Parameters
data – Receive the data format in pd.DataFrame format. ImageID is an essential element for data.
inference – Receive the inference format in the form of pd.DataFrame. ImageID and ClassProbability are essential for inferences.
- get_sample(method: str, k: int, n: int = 3) pandas.core.frame.DataFrame[source]
A function that extracts sample data and returns it.
- Parameters
method –
‘topk’ - It extracts the k sample data with the highest uncertainty.
’lowk’ - It extracts the k sample data with the lowest uncertainty.
’randomk’ - Extract and return random k sample data.
k – number of sample data
n – Parameters to be used in the randtopk method, Variable to first extract data of multiple n of k.
- Returns
pd.DataFrame
- Return type
Extracted sample data
- _get_sample_mixed(method: str, k: int, n: int = 3) pandas.core.frame.DataFrame[source]
A function that extracts sample data and returns it.
- Parameters
method –
‘mixk’: Return top-k and low-k halves based on uncertainty.
’randomtopk’: Randomly extract n*k and return k with high uncertainty.
k – number of sample data
n – Number to extract n * k from total data according to n, and top-k from it
- Returns
pd.DataFrame
- Return type
Extracted sample data