The system according to the human gesture is determined whether the person sleeping in FIG.
A data crawling
Take pictures from Baidu climb to more than 1000 images of the body non-sleeping and sleeping posture pictures
Second, data extraction
A picture of each key point is detected, the data is stored to a csv file, and add the flag 0 (no sleep) and 1 (sleep). Since the decision to sleep posture is the key point
key_points = ['top_head', 'neck', 'left_shoulder', 'right_shoulder', 'left_elbow', 'right_elbow', 'left_wrist', 'right_wrist', 'left_hip', 'right_hip', 'left_knee', 'right_knee', 'left_ankle', 'right_ankle'],所以,只需获取这些关键点。
Third, data cleaning
Remove duplicate data containing a large amount of data with null values
Fourth, machine learning, modeling
Use sklearn random forest regression models for modeling
Fifth, the model evaluation
Confusion matrix:
[[125 0]
[ 13 11]]