Under carefully read today, if the class read the original Image rewrite the object image data directly read something does not seem so simple. Previous here .
For the wisdom of others just mess things are what the network structure, run the model, I do ye up all the details? All pit? ? ?
My feeling is misguided.
I thought of a way of lazy, I really do not want to write these things from scratch, time-consuming and thankless, or taken advantage of.
Since it takes an Image object I give it wants this stuff, why bother? ?
postscript:
Ffmpeg found that the use of divided frames than cv2 frame Cap 3
Finally ran through a direct test video, but what use is it? ?
I was on an action directly test? Or training a network?
Or feature extraction using cluster? ? ?
If someone fly with me nice ah.
Not tested. The election of the first direction, read this stuff out more than two weeks, I feel like I'm a rookie. . . . . . . . . .