def face_detect(image, config='haarcascade_frontalface_default'):
""" Detects whether in the given image (PIL image) there are any faces
or not using the OpenCV Haar cascade detector. """
# Save the PIL image to disk and re-load it into IplImage format for
# OpenCV. I know this is a terrible hack but we can take care of this
# as an in-memory conversion later.
from files import random_file_name
temp_file = random_file_name('jpg')
image.save(temp_file)
opencv_image = cv.LoadImage(temp_file, cv.CV_LOAD_IMAGE_COLOR)
return _face_detect(opencv_image, config)
def voting_face_detect(image, delta=20):
""" Runs all detectors on an image and returns the most voted one.
"""
configs = [config for config in listdir("%s/resources/opencv_detectors" %
PROJECT_PATH)]
from files import random_file_name
temp_file = random_file_name('jpg')
image.save(temp_file)
opencv_image = cv.LoadImage(temp_file, cv.CV_LOAD_IMAGE_COLOR)
results = [_face_detect(opencv_image, config) for config in configs]
logging.debug('found results %s' % zip(configs, results))
results = filter(lambda x: len(x) != 0, results)
all_results = []
for r in results:
all_results.extend([r[0]])
votes = [0] * len(all_results)
logging.debug('found nb results %s' % len(all_results))
if len(votes) == 0:
return []
if len(configs) == 1:
return all_results[0]
for i in range(len(all_results)):
for j in range(i + 1, len(all_results)):
if similar(all_results[i][0], all_results[j][0], delta):
votes[i] += 1
logging.debug('found votes %s' % votes)
max_votes = max(votes)
if max_votes < 1:
return []
return results[votes.index(max_votes)]
