follow a more common format

ipazc / mtcnn

MTCNN face detection implementation for TensorFlow, as a PIP package.

MIT License

2.23k stars 527 forks source link

follow a more common format #105

Closed ma7555 closed 3 weeks ago

ma7555 commented 3 years ago

return xmin, ymin, xmax, ymax instead of x, y, width, heigth

MattyB95 commented 3 years ago

@ma7555 Why have you removed the - x and - y from the width and height. That is there as part of the fix for Pull Request #94 unless something has happened since that I am not aware of.

Also, I'm not sure how much I agree with this change suggestion; these names suggest coordinates which is not the case for the width and height. However, I haven't looked at this project in detail in a little while, just my two cents worth.

ma7555 commented 3 years ago

Hello @MattyB95,

The pull request totally replaces the width and height with xmax and ymax points so the box in returned as 4 points, not 2 points + width + height. I don't understand the -x and -y you are refering to but the whole line has been omitted.

This format is what is returned by most recent face detection models like mobileface, retinaface, etc..