In pretrained.cc, how to get the bounding boxes?

[andviane]

In pretrained.cc, is it possible to obtain the bounding boxes and not just the matching scores? Or is this functionality normally implemented already on the top of the Caffe2?

[alexge233]

@andviane Bounding Boxes are usually a computer vision task, not a deep learning task. I know there are DL-based face detectors which provide bounding boxes, but it's a lot easier if you just use OpenCV or DLib (for faces). It can be done with Caffe2/PyTorch/etc but you'd have to train the classifier to return pixel values or coordinates I'd suppose.