√ python 录屏 + 脸部识别 python屏幕录制
√ 先跑通截屏的code，之后将脸部识别融入进去
problems: 两个人的人脸数据坐标混乱。不能定向追踪某个人。
solution: 屏幕一分为二，录屏的坐标固定，限制在某一坐标内的人脸就是固定的faceid
(TBD) 本地拿到两组人脸数据（中心点）判断是否同步
需要的库依赖:
dependency:
pynput
Pillow
python-opencv
the x and y to seperate coordinates of two faces.
how to study the sync
save the result to local file??
or save the data into arrays and do real-time analysis??
GUI: two little window with energy bar and link to the sync result
locate the precise x and y of the windows.
弹窗
first energy bar: x:800 y:660 w:66 h:200
second energy bar: x:1700 y:660 w:66 h:200
优化输出数据，如果没检测到和上一帧保持一样 判断相关性
use a queue to save 15 frames
calculate the standard deviation of the 15 number
use the result as an indication whether the user has nodded
works to be done:
- split the window (now we have two bars in the same window)
- move the window to the right place (the information of the position is listed above)
- make sure the windows are always on top
- find out other ways to visualize the output (optional)
- find volunteers?
- start the report
The layout is far from good, need to change the place of the window
need to at least change the color
or the position
User study
Things to prepare: A topic, environment set up, light, camera, questionnaire
will do some experiment today, successful or not, will introduce to others and collect some data.
- make a online survey, which is easier to gather the results.
- try to compile the python code into a package so people can download the package and use the design without downloading the python libs.
- already start to write the paper and the report.