def labeling(etapa, modelo1, modelo2, modelo3, arquitectura1, arquitectura2,
arquitectura3, datos, pipeline, iteracion, models_info):
"""
Labels samples using models
Args:
etapa (str): Word 1 to compare
modelo1 (bytes): Top1 Model
modelo2 (bytes): Top2 Model
modelo3 (bytes): Top3 Model
arquitectura1 (str): Top1 Arch
arquitectura2 (str): Top2 Arch
arquitectura3 (str): Top3 Arch
EL (list): Enlarge labeled samples
LC (list): Low confidence samples
datos (df): Dataframe with data
pipeline (dict): General config
iteracion (int): Semi-supervised Stage
models_info (list): Info about models
Returns:
EL (list): Enlarge labeled samples updated
LC (list): Low confidence samples updated
EL_iter ():
LC_iter ():
"""
etiquetados_EL = 0
etiquetados_LC = 0
EL_iter = []
LC_iter = []
predicciones = []
predicciones_logs = []
train_generator_arch1, batchset_generator_arch1, STEP_SIZE_BATCH_arch1 = generadores(
etapa, arquitectura1, datos, pipeline, True, iteracion, models_info)
train_generator_arch2, batchset_generator_arch2, STEP_SIZE_BATCH_arch2 = generadores(
etapa, arquitectura2, datos, pipeline, True, iteracion, models_info)
train_generator_arch3, batchset_generator_arch3, STEP_SIZE_BATCH_arch3 = generadores(
etapa, arquitectura3, datos, pipeline, True, iteracion, models_info)
df1 = evaluar(modelo1, train_generator_arch1, batchset_generator_arch1,
STEP_SIZE_BATCH_arch1)
df2 = evaluar(modelo2, train_generator_arch2, batchset_generator_arch2,
STEP_SIZE_BATCH_arch2)
df3 = evaluar(modelo3, train_generator_arch3, batchset_generator_arch3,
STEP_SIZE_BATCH_arch3)
for i in range(len(df1)):
arch_scores = {}
arch_scores[arquitectura1] = df1['Max_Probability'][i]
arch_scores[arquitectura2] = df2['Max_Probability'][i]
arch_scores[arquitectura3] = df3['Max_Probability'][i]
c1 = (df1['Predictions'][i] == df2['Predictions'][i])
c2 = (df1['Predictions'][i] == df3['Predictions'][i])
c3 = (df2['Predictions'][i] == df3['Predictions'][i])
p1 = (df1['Max_Probability'][i] > pipeline["ssl_threshold"])
p2 = (df2['Max_Probability'][i] > pipeline["ssl_threshold"])
p3 = (df3['Max_Probability'][i] > pipeline["ssl_threshold"])
if c1 and c2 and p1 and p2 and p3:
datos["EL"].append(
[df1['Filename'][i], df1['Predictions'][i], arch_scores])
predicciones.append([df1['Filename'][i], df1['Predictions'][i]])
selected = df1['Predictions'][i]
prob_selected = df1["Max_Probability"][i]
predicciones_logs.append([
df1['Filename'][i], selected, prob_selected, "EL",
df1['Predictions'][i], df1['Max_Probability'][i],
df2["Predictions"][i], df2['Max_Probability'][i],
df3["Predictions"][i], df3['Max_Probability'][i]
])
EL_iter.append(
[df1['Filename'][i], df1['Predictions'][i], arch_scores])
etiquetados_EL += 1
else:
datos["LC"].append(
[df1['Filename'][i], df1['Predictions'][i], arch_scores])
predicciones.append([df1['Filename'][i], df1['Predictions'][i]])
selected = df1['Predictions'][i]
prob_selected = df1["Max_Probability"][i]
predicciones_logs.append([
df1['Filename'][i], selected, prob_selected, "LC",
df1['Predictions'][i], df1['Max_Probability'][i],
df2["Predictions"][i], df2['Max_Probability'][i],
df3["Predictions"][i], df3['Max_Probability'][i]
])
LC_iter.append(
[df1['Filename'][i], df1['Predictions'][i], arch_scores])
etiquetados_LC += 1
print('etiquetados EL {} LC {}'.format(etiquetados_EL, etiquetados_LC))
import pandas as pd
df_logs = pd.DataFrame(predicciones_logs,
columns=[
"y_true", "y_pred", "prob_pred", "type",
"pred1", "prob1", "pred2", "prob2", "pred3",
"prob3"
])
df_logs["y_true"] = df_logs['y_true'].apply(lambda x: x.split('/')[-2])
print("LABELING LOGS")
print(df_logs)
df_EL_logs = df_logs[(df_logs["type"] == "EL")].copy()
df_LC_logs = df_logs[(df_logs["type"] == "LC")].copy()
y_true = df_EL_logs['y_true'].values.tolist()
y_pred = df_EL_logs['y_pred'].values.tolist()
from sklearn.metrics import accuracy_score
EL_accu = accuracy_score(y_true, y_pred)
print("EL_ACCU: ", EL_accu)
y_true = df_LC_logs['y_true'].values.tolist()
y_pred = df_LC_logs['y_pred'].values.tolist()
from sklearn.metrics import accuracy_score
LC_accu = accuracy_score(y_true, y_pred)
print("LC_ACCU: ", LC_accu)
return datos, EL_iter, LC_iter, [df1, df2, df3]
def labeling(etapa, modelo1, modelo2, modelo3, arquitectura1, arquitectura2,
arquitectura3, datos, pipeline, iteracion, models_info):
"""
Labels samples using models
Args:
etapa (str): Word 1 to compare
modelo1 (bytes): Top1 Model
modelo2 (bytes): Top2 Model
modelo3 (bytes): Top3 Model
arquitectura1 (str): Top1 Arch
arquitectura2 (str): Top2 Arch
arquitectura3 (str): Top3 Arch
EL (list): Enlarge labeled samples
LC (list): Low confidence samples
datos (df): Dataframe with data
pipeline (dict): General config
iteracion (int): Semi-supervised Stage
models_info (list): Info about models
Returns:
EL (list): Enlarge labeled samples updated
LC (list): Low confidence samples updated
EL_iter ():
LC_iter ():
"""
etiquetados_EL = 0
etiquetados_LC = 0
EL_iter = []
LC_iter = []
train_generator_arch1, batchset_generator_arch1, STEP_SIZE_BATCH_arch1 = generadores(
etapa, arquitectura1, datos, pipeline, True, iteracion, models_info)
train_generator_arch2, batchset_generator_arch2, STEP_SIZE_BATCH_arch2 = generadores(
etapa, arquitectura2, datos, pipeline, True, iteracion, models_info)
train_generator_arch3, batchset_generator_arch3, STEP_SIZE_BATCH_arch3 = generadores(
etapa, arquitectura3, datos, pipeline, True, iteracion, models_info)
df1 = evaluar(modelo1, train_generator_arch1, batchset_generator_arch1,
STEP_SIZE_BATCH_arch1)
df2 = evaluar(modelo2, train_generator_arch2, batchset_generator_arch2,
STEP_SIZE_BATCH_arch2)
df3 = evaluar(modelo3, train_generator_arch3, batchset_generator_arch3,
STEP_SIZE_BATCH_arch3)
for i in range(len(df1)):
arch_scores = {}
arch_scores[arquitectura1] = df1['Max_Probability'][i]
arch_scores[arquitectura2] = df2['Max_Probability'][i]
arch_scores[arquitectura3] = df3['Max_Probability'][i]
c1 = (df1['Predictions'][i] == df2['Predictions'][i]) and (
df1['Max_Probability'][i] > pipeline["ssl_threshold"]) and (
df2['Max_Probability'][i] > pipeline["ssl_threshold"])
c2 = (df1['Predictions'][i] == df3['Predictions'][i]) and (
df1['Max_Probability'][i] > pipeline["ssl_threshold"]) and (
df3['Max_Probability'][i] > pipeline["ssl_threshold"])
c3 = (df2['Predictions'][i] == df3['Predictions'][i]) and (
df2['Max_Probability'][i] > pipeline["ssl_threshold"]) and (
df3['Max_Probability'][i] > pipeline["ssl_threshold"])
if c1 and c2 and c3:
datos["EL"].append(
[df1['Filename'][i], df1['Predictions'][i], arch_scores])
EL_iter.append(
[df1['Filename'][i], df1['Predictions'][i], arch_scores])
etiquetados_EL += 1
else:
datos["LC"].append(
[df1['Filename'][i], df1['Predictions'][i], arch_scores])
LC_iter.append(
[df1['Filename'][i], df1['Predictions'][i], arch_scores])
etiquetados_LC += 1
print('etiquetados EL {} LC {}'.format(etiquetados_EL, etiquetados_LC))
return datos, EL_iter, LC_iter
def labeling_v2(etapa, modelo1, modelo2, modelo3, arquitectura1, arquitectura2,
arquitectura3, datos, pipeline, iteracion, models_info):
predicciones = []
predicciones_logs = []
etiquetados_EL = 0
etiquetados_LC = 0
EL_iter = []
LC_iter = []
train_generator_arch1, batchset_generator_arch1, STEP_SIZE_BATCH_arch1 = generadores(
etapa, arquitectura1, datos, pipeline, True, iteracion, models_info)
train_generator_arch2, batchset_generator_arch2, STEP_SIZE_BATCH_arch2 = generadores(
etapa, arquitectura2, datos, pipeline, True, iteracion, models_info)
train_generator_arch3, batchset_generator_arch3, STEP_SIZE_BATCH_arch3 = generadores(
etapa, arquitectura3, datos, pipeline, True, iteracion, models_info)
df1 = evaluar(modelo1, train_generator_arch1, batchset_generator_arch1,
STEP_SIZE_BATCH_arch1)
df2 = evaluar(modelo2, train_generator_arch2, batchset_generator_arch2,
STEP_SIZE_BATCH_arch2)
df3 = evaluar(modelo3, train_generator_arch3, batchset_generator_arch3,
STEP_SIZE_BATCH_arch3)
for i in range(len(df1)):
arch_scores = {}
arch_scores[arquitectura1] = df1['Max_Probability'][i]
arch_scores[arquitectura2] = df2['Max_Probability'][i]
arch_scores[arquitectura3] = df3['Max_Probability'][i]
c1 = (df1['Predictions'][i] == df2['Predictions'][i])
c2 = (df1['Predictions'][i] == df3['Predictions'][i])
c3 = (df2['Predictions'][i] == df3['Predictions'][i])
if c1 and c2:
datos["EL"].append(
[df1['Filename'][i], df1['Predictions'][i], arch_scores])
predicciones.append([df1['Filename'][i], df1['Predictions'][i]])
selected = df1['Predictions'][i]
prob_selected = df1["Max_Probability"][i]
predicciones_logs.append([
df1['Filename'][i], selected, prob_selected, "EL",
df1['Predictions'][i], df1['Max_Probability'][i],
df2["Predictions"][i], df2['Max_Probability'][i],
df3["Predictions"][i], df3['Max_Probability'][i]
])
EL_iter.append(
[df1['Filename'][i], df1['Predictions'][i], arch_scores])
etiquetados_EL += 1
elif c1 or c2:
datos["EL"].append(
[df1['Filename'][i], df1['Predictions'][i], arch_scores])
predicciones.append([df1['Filename'][i], df1['Predictions'][i]])
selected = df1['Predictions'][i]
prob_selected = df1["Max_Probability"][i]
predicciones_logs.append([
df1['Filename'][i], selected, prob_selected, "EL",
df1['Predictions'][i], df1['Max_Probability'][i],
df2["Predictions"][i], df2['Max_Probability'][i],
df3["Predictions"][i], df3['Max_Probability'][i]
])
EL_iter.append(
[df1['Filename'][i], df1['Predictions'][i], arch_scores])
etiquetados_EL += 1
elif c3:
datos["EL"].append(
[df1['Filename'][i], df1['Predictions'][i], arch_scores])
predicciones.append([df2['Filename'][i], df2['Predictions'][i]])
selected = df2['Predictions'][i]
prob_selected = df2["Max_Probability"][i]
predicciones_logs.append([
df1['Filename'][i], selected, prob_selected, "EL",
df1['Predictions'][i], df1['Max_Probability'][i],
df2["Predictions"][i], df2['Max_Probability'][i],
df3["Predictions"][i], df3['Max_Probability'][i]
])
EL_iter.append(
[df1['Filename'][i], df1['Predictions'][i], arch_scores])
etiquetados_EL += 1
else:
datos["LC"].append(
[df1['Filename'][i], df1['Predictions'][i], arch_scores])
LC_iter.append(
[df1['Filename'][i], df1['Predictions'][i], arch_scores])
etiquetados_LC += 1
predicciones.append([df1['Filename'][i], df1['Predictions'][i]])
selected = df1['Predictions'][i]
prob_selected = df1["Max_Probability"][i]
predicciones_logs.append([
df1['Filename'][i], selected, prob_selected, "LC",
df1['Predictions'][i], df1['Max_Probability'][i],
df2["Predictions"][i], df2['Max_Probability'][i],
df3["Predictions"][i], df3['Max_Probability'][i]
])
print('etiquetados EL {} LC {}'.format(etiquetados_EL, etiquetados_LC))
import pandas as pd
df_logs = pd.DataFrame(predicciones_logs,
columns=[
"y_true", "y_pred", "prob_pred", "type",
"pred1", "prob1", "pred2", "prob2", "pred3",
"prob3"
])
df_logs["y_true"] = df_logs['y_true'].apply(lambda x: x.split('/')[-2])
print("LABELING LOGS")
print(df_logs)
df_EL_logs = df_logs[(df_logs["type"] == "EL")].copy()
df_LC_logs = df_logs[(df_logs["type"] == "LC")].copy()
y_true = df_EL_logs['y_true'].values.tolist()
y_pred = df_EL_logs['y_pred'].values.tolist()
from sklearn.metrics import accuracy_score
EL_accu = accuracy_score(y_true, y_pred)
print("EL_ACCU: ", EL_accu)
y_true = df_LC_logs['y_true'].values.tolist()
y_pred = df_LC_logs['y_pred'].values.tolist()
from sklearn.metrics import accuracy_score
LC_accu = accuracy_score(y_true, y_pred)
print("LC_ACCU: ", LC_accu)
return datos, EL_iter, LC_iter, [df1, df2, df3]