CODEWITHSUNDEEP
pythonopencvdistanceface-detectionmeasure
T.R.
T.R.

Reputation: 33

Real width of detected face

I've been researching like forever, but couldn't find an answer. I'm using OpenCV to detect faces, now I want to calculate the distance to the face. When I detect a face, I get a matofrect (which I can visualize with a rectangle). Pretty clear so far. But now: how do I get the width of the rectangle in the real world? There has to be some average values that represent the width of the human face. If I have that value (in inch, mm or whatever), I can calculate the distance using real width, pixel width and focal length. Please, can anyone help me?

Note: I'm comparing the "simple" rectangle solution against a Facemark based distance measuring solution, so no landmark based answers. I just need the damn average face / matofrectwidth :D

Thank you so much!

Upvotes: 1

Views: 336

Answers (1)

Andrew T
Andrew T

Reputation: 193

OpenCV's facial recognition is slightly larger than a face, therefore an average face may not be helpful. Instead, just take a picture of a face at different distances from the camera and record the distance from the camera along with the pixel width of the face for several distances. After plotting the two variables on a graph, use a trendline to come up with a predictive model.

Upvotes: 2

Related Questions