Scaling in correct_yolo_boxes seems not required

[billysz]

Hi,

Since (xmin, xmax, ymin, ymax) are already normalized values, I think no need to re-scale them any more, just applying to the original image should be fine, as following:
boxes[i].xmin = int((boxes[i].xmin) * image_w)
boxes[i].xmax = int((boxes[i].xmax) * image_w)
boxes[i].ymin = int((boxes[i].ymin) * image_h)
boxes[i].ymax = int((boxes[i].ymax) * image_h)

Regards,