How to handle no face in image

[Joaquin-aliaga]

Hi, I'm developing a face verification system, so I'm using mtcnn to detect faces and then do the verification step.

The code I have is something like:

face_detect, prob = mtcnn.detect(img)
if (face_detect is not None):
    face_cropped = mtcnn(img)
   ...
else:
   return None

However, I'm wondering if there is a way to handle the "no face in image" case without use mtcnn twice, because I assume mtcnn.detect() and mtcnn() use almost the same code.

If I just use face_cropped = mtcnn(img) and img has no face, I get an error.

line 408, in select_boxes
    if len(boxes) == 0:
TypeError: len() of unsized object

Thanks in advance

[Joaquin-aliaga]

@Joaquin-aliaga I've already find a solution for not using mtccn twice and post here if anyone faces the same issue. Since mtcnn.detect() returns bounding boxes and probs, just use:

bbx, prob = mtcnn.detect(img)
if (bbx is not None):
    face_cropped = mtcnn.extract(img,bbx,None)
   ...
else:
   return None

In that way you don't have to pass your image twice through the CNN.

However it would be nice if mtcnn() returns None when no face was detected.

[justDoIt1314]

The problem has been solved.

if type(boxes) == numpy.ndarray:
    for bbox in boxes: