Bounding boxes returned without detected face image in dlib python

I'm trying to detect multiple faces in a picture using the deepface library with dlib as the backend detector. I'm using the DlibWrapper.py from the deeepface library and i have the following issue: In some cases, the detector returns the bounding box coordinates but doesn't return the detected face image detected face-box coordinates.

I was wondering if this bug occurs because of the negative values of some coordinates of the bounding boxes but i figured out that was not the case, as the negative values are features, not bugs. Here is the DlibWrapper from the deepface library.

Solved!There are edge cases where original rectangle is partially outside the image window. That happens with dlib. So, instead of

• detected_face = img[top:bottom, left:right],

the detected face should be

• detected_face = img[max(0, top): min(bottom, img_height), max(0, left): min(right, img_width)]