1
$\begingroup$

I'd like to have my undergrad machine learning students have the option of doing a face detection project using neural networks (constructed by the students using Keras). The algorithm should ideally be simple but able to detect faces in a group photo of our class. From what I've seen so far, implementing YOLO or R-CNN looks too complicated, and I think a simpler approach should be possible because YOLO and R-CNN are near state of the art and they do more than just face detection.

Any recommendations for a relatively simple face detection algorithm that could be used?

$\endgroup$
1
2
$\begingroup$

Relatively trivial approach to object (face) detection is to train a (convolutional) neural network to perform binary classification (face vs. no face) on image patches. Then you can pass patches of the tested image in a sliding window fashion (using certain stride for improved speed) to the network and see if it has detected a face in that region of the tested image. Finally, to handle various sizes of faces, you can process the tested image on several different scales.

$\endgroup$
2
  • $\begingroup$ Thanks. What do you think would be a good training dataset to use to train the binary classifier (face vs. no face) ? $\endgroup$
    – littleO
    Apr 1 '19 at 15:55
  • $\begingroup$ quick Google search produces e.g. this one: kaggle.com/dataturks/face-detection-in-images $\endgroup$ Apr 2 '19 at 11:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.