[note]Some sample code of python#3

筆記程式方便查詢

Removing duplicates in the lists

>>> t = [1, 2, 3, 1, 2, 5, 6, 7, 8]
>>> t
[1, 2, 3, 1, 2, 5, 6, 7, 8]
>>> list(set(t))
[1, 2, 3, 5, 6, 7, 8]

Evaluation Metrics

# print("  FN    FP   TP     pre   acc   rec   f1")
#print(FN, FP, TP, FN+FP+TP+TF)
precision = TP / (TP + FP)
print(f"precision: {precision:4.2f}")
accuracy = (TP + TN)/(TP + TN + FP + FN)
print(f"accuracy: {recall:4.2f}")
recall = TP / (TP + FN)
print(f"recall: {recall:4.2f}")
f1_score = 2 * precision * recall / (precision + recall)
print(f"f1_score: {f1_score:4.2f}")
# print(f"{FN:6.2f}{FP:6.2f}{TP:6.2f}", end="")
# print(f"{precision:6.2f}{accuracy:6.2f}{recall:6.2f}{f1_score:6.2f}")

Confusion Matrix — Get Items FP/FN/TP/TN — Python

import sklearn    
from sklearn.metrics import confusion_matrix
actual = [1, -1, 1, 1, -1, 1]
predicted = [1, 1, 1, -1, -1, 1]
confusion_matrix(actual, predicted)
>>>
array([[1, 1],
       [1, 3]])
c = confusion_matrix(actual, predicted)
TN, FP, FN, TP = c[0][0], c[0][1], c[1][0],c[1][1]

Set a threshold of IoU to determine if the object detection is valid or not

Let’s say you set IoU to 0.5, in that case (From Renu Khandelwal)

• True Positive(TP): if IoU ≥0.5, classify the object detection as TP
• False Positive(FP): if Iou <0.5, then it is a wrong detection and classify it as FP
• True Negative (TN): TN is every part of the image where we did not predict an object. This metrics is not useful for object detection, hence we ignore TN.
• False Negative(FN): When a ground truth is present in the image and model failed to detect the object, classify it as FN.

Set IoU threshold value to 0.5 or greater(0.5, 0.75. 0.9 or 0.95…).

Reference

Python:刪除重複值

Evaluating performance of an object detection model

Precision-Recall - scikit-learn 0.24.0 documentation