def print_stats(self, ground_truths, predictions, name):
cm = confusion_matrix(ground_truths, predictions, self.classes)
print("###" + name + " Results###\n")
# print_cm(cm, self.classes)
# print("\n\n")
print_cm(np.around(cm.astype('float') / cm.sum(axis=1)[:, np.newaxis], decimals=3), self.classes)
print("\n\n")
print("Accuracy score: ", accuracy_score(ground_truths, predictions), "\n\n")
# print("Report")
# print(classification_report(ground_truths, predictions))
print("#################################################################end###\n\n\n")
def model_eval(self, mode="sample"):
if mode not in {"sample", "clip"}:
raise Exception("Evaluation mode not allowed.")
predictions = []
ground_truths = []
stats = []
if mode == "sample":
files = glob.glob("/user/vlongobardi/late_feature/" +
self.feature_name + "/Val/*/*")
val_gen = self.data_gen(files, 1, "eval")
for batch in val_gen:
ground_truth = self.lb.inverse_transform(batch[1])[0]
pred = self.lb.inverse_transform(self.model.predict(
batch[0]))[0]
predictions.append(pred)
ground_truths.append(ground_truth)
else:
files = glob.glob(
"/user/vlongobardi/late_feature/emobase2010_full_wav/Val/*/*")
files = [
file.replace("emobase2010_full_wav",
self.feature_name).split(".")[0] for file in files
]
for file in files:
ground_truths.append(file.split("/")[-2])
predictions.append(self.clip_classification(file))
cm = confusion_matrix(ground_truths, predictions, self.classes)
stats.append(cm)
stats.append(
np.around(cm.astype('float') / cm.sum(axis=1)[:, np.newaxis],
decimals=2))
stats.append(accuracy_score(ground_truths, predictions))
stats.append(classification_report(ground_truths, predictions))
print("###Results###")
for index, elem in enumerate(stats):
if index < 2:
print_cm(elem, self.classes)
elif index == 2:
print("Accuracy score: ", elem)
else:
print("Report")
print(elem)
print("\n\n")