def main(mtcnn_model_dir, save_dir=None):
if save_dir is None:
save_dir = './result_after_align'
if not osp.exists(save_dir):
print('mkdir for aligned faces, aligned root dir: ', save_dir)
os.makedirs(save_dir)
aligner = MtcnnAligner(mtcnn_model_dir, False)
fp = open('./dataset-gender-all.json', 'r')
all_lines = fp.readlines()
count = 25499
for line in all_lines[25499:]:
err_msg = ''
count = count + 1
print count
data = json.loads(line)
image_url = data["url"]
image_bbox = data["label"]["detect"]["general_d"]["bbox"]
#"label":{"detect":{"general_d":{"bbox":[{"class":"male","pts":
print image_url
print image_bbox
#print len(image_bbox)
for idx in range(len(image_bbox)):
if image_bbox[idx] == []:
continue
else:
img = io.imread(image_url)
if len(img.shape) != 3:
continue
else:
file_idx = "%06d" % count
print file_idx
filename = str(file_idx) + '_' + str(idx) + '.json'
fp_rlt = open(osp.join(save_dir, filename), 'w')
item = {}
print('===> Processing image: ' + filename)
per_bbox = image_bbox[idx]["pts"]
print "bbox:", per_bbox
GT_RECT = [
per_bbox[0][0], per_bbox[0][1], per_bbox[2][0],
per_bbox[2][1]
]
print "GT:", GT_RECT
boxes, points = aligner.align_face(img, [GT_RECT])
box = boxes[0]
pts = points[0]
item['url'] = image_url
item['class'] = image_bbox[idx]["class"]
tmp = {'rect': GT_RECT, 'pts': pts}
item['detect'] = tmp
print "item:", item
json.dump(item, fp_rlt, indent=4)
fp_rlt.close()
class FaceAligner:
def __init__(self, caffe_model_path=None):
self.aligner = None
if caffe_model_path:
self.aligner = MtcnnAligner(caffe_model_path)
def align_face(self, img, face_rects):
if isinstance(img, str):
img = cv2.imread(img)
regressed_rects, facial_points = self.aligner.align_face(
img, face_rects)
return (regressed_rects, facial_points)
def get_face_chips(self, img, face_rects, facial_points=None):
if facial_points is None:
if self.aligner is None:
raise Exception('FaceAligner.aligner is not initialized')
rects, facial_points = self.aligner.align_face(img, face_rects)
reference_5pts = None
output_size = (96, 112) # (w, h) not (h,w)
face_chips = []
for facial_5pts in facial_points:
facial_5pts = np.reshape(facial_5pts, (2, -1))
dst_img = warp_and_crop_face(img, facial_5pts, reference_5pts,
output_size)
face_chips.append(dst_img)
return face_chips
def main(img_path, face_rects, save_dir=None, save_img=True, show_img=True):
if save_dir is None:
save_dir = './fa_test_rlt'
save_json = 'mtcnn_align_rlt.json'
caffe_model_path = "../../model"
if not osp.exists(save_dir):
os.makedirs(save_dir)
fp_rlt = open(osp.join(save_dir, save_json), 'w')
results = []
print '===> Processing image: ' + img_path
img = cv2.imread(img_path)
aligner = MtcnnAligner(caffe_model_path, False)
rlt = {}
rlt["filename"] = img_path
rlt["faces"] = []
rlt['face_count'] = 0
t1 = time.clock()
bboxes, points = aligner.align_face(img, face_rects)
t2 = time.clock()
n_boxes = len(face_rects)
print(
"-->Alignment cost %f seconds, processed %d face rects, avg time: %f seconds"
% ((t2 - t1), n_boxes, (t2 - t1) / n_boxes))
if bboxes is not None and len(bboxes) > 0:
for (box, pts) in zip(bboxes, points):
# box = box.tolist()
# pts = pts.tolist()
tmp = {'rect': box[0:4], 'score': box[4], 'pts': pts}
rlt['faces'].append(tmp)
rlt['face_count'] = len(bboxes)
rlt['message'] = 'success'
results.append(rlt)
if save_img or show_img:
draw_faces(img, bboxes, points)
if save_img:
save_name = osp.join(save_dir, osp.basename(img_path))
cv2.imwrite(save_name, img)
if show_img:
cv2.imshow('img', img)
cv2.waitKey(0)
cv2.destroyAllWindows()
json.dump(results, fp_rlt, indent=4)
fp_rlt.close()
def main(face_json_file, mtcnn_model_dir, save_dir=None):
if not osp.exists(save_dir):
print('mkdir for aligned faces, aligned root dir: ', save_dir)
os.makedirs(save_dir)
if not osp.exists(aligned_save_dir):
print('mkdir for aligned faces, aligned root dir: ', save_dir)
os.makedirs(aligned_save_dir)
aligner = MtcnnAligner(mtcnn_model_dir, False)
#root_dir = '/disk2/du/face-asian/'
#path_walk = os.walk(root_dir)
fp = open('./face_asian_list.txt','r')
all_lines = fp.readlines()
print len(all_lines)
count = 180001
for line in all_lines[180000:360000]:
print line
#for root,dirs,files in path_walk:
err_msg = ''
print count
count = count + 1
#data = json.loads(line)
#image_url = data["url"]
#print image_url
overlap_thresh_0 = overlap_thresh
#save_path = osp.join(data[u'url'].split('/')[-5],data[u'url'].split('/')[-4],data[u'url'].split('/')[-3],data[u'url'].split('/')[-2])
save_path = line.split('/')[-2]
print osp.join(save_dir, save_path)
isExists1 = osp.exists(osp.join(save_dir,save_path))
if not isExists1:
os.makedirs(osp.join(save_dir,save_path))
#save_img_fn = osp.join(data[u'url'].split('/')[-5],data[u'url'].split('/')[-4],data[u'url'].split('/')[-3],data[u'url'].split('/')[-2])
save_img_fn = line.split('/')[-2]
save_fn =aligned_save_dir + '/' + save_img_fn + '/' + line.split('/')[-1]
print save_fn
isExists2 = osp.exists(osp.join(aligned_save_dir,save_img_fn))
if not isExists2:
os.makedirs(osp.join(aligned_save_dir,save_img_fn))
filename = line.split('/')[-1][:-6] + '.json'
print filename
fp_rlt = open(osp.join(save_dir,save_path,filename), 'w')
item = {}
print('===> Processing image: ' + save_path + '/' + filename)
#imageBGR = io.imread(data[u'url'])
#image = cv2.cvtColor(imageBGR, cv2.COLOR_BGR2RGB)
image = cv2.imread(line[:-2])
print image.shape
GT_RECT = [0,0,256,256]
boxes, points = aligner.align_face(image, [GT_RECT])
box = boxes[0]
pts = points[0]
facial5points = np.reshape(points, (2, -1))
dst_img = warp_and_crop_face(image, facial5points, reference_5pts, output_size)
cv2.imwrite(save_fn, dst_img)
tmp = {'rect': box[0:4],
'score': box[4],
'pts': pts
}
item['faces'] = tmp
#item['id'] = data[u'url'].split('/')[-3]
item['shape'] = image.shape
json.dump(item, fp_rlt, indent=4)
fp_rlt.close()
def __init__(self, caffe_model_path=None):
self.aligner = None
if caffe_model_path:
self.aligner = MtcnnAligner(caffe_model_path)
def main(args):
rect_root_dir = args.rect_root_dir
save_dir = args.save_dir
mtcnn_model_dir = args.mtcnn_model_dir
img_root_dir = args.image_root_dir
gpu_id = args.gpu_id
MAX_LINE = 500
rect_list = get_rect_list(rect_root_dir)
print('all %d images' % (len(rect_list)))
if not save_dir:
save_dir = './aligned_root_dir'
if not osp.exists(save_dir):
print('makedirs for aligned root dir: ', save_dir)
os.makedirs(save_dir)
save_aligned_dir = osp.join(save_dir, 'aligned_imgs')
if not osp.exists(save_aligned_dir):
print('makedirs for aligned/cropped face imgs: ', save_dir)
os.makedirs(save_aligned_dir)
save_rects_dir = osp.join(save_dir, 'face_rects')
if not osp.exists(save_rects_dir):
print('makedirs for face rects/landmarks: ', save_rects_dir)
os.makedirs(save_rects_dir)
aligner = MtcnnAligner(mtcnn_model_dir, True, gpu_id=gpu_id)
for i in range(len(rect_list)):
with open(osp.join(rect_root_dir, rect_list[i]), 'r') as f:
#lines = f.readlines()
contents = json.load(f)
img_name = contents['filename']
face_count = contents['face_count']
if face_count == 0:
print('not detect face: %s' % img_name)
continue
img = cv2.imread(img_name)
img_h = img.shape[0]
img_w = img.shape[1]
max_line = 0
resize_scale = 0
new_img_h = img_h
new_img_w = img_w
if img_h >= img_w:
max_line = img_h
if max_line > MAX_LINE:
resize_scale = max_line / MAX_LINE
new_img_h = MAX_LINE
new_img_w = img_w / resize_scale
if img_w > img_h:
max_line = img_w
if max_line > MAX_LINE:
resize_scale = max_line / MAX_LINE
new_img_w = MAX_LINE
new_img_h = img_h / resize_scale
print 'image.shape:', img.shape
img = cv2.resize(img, (new_img_w, new_img_h))
img_center_x = img_w / 2
img_center_y = img_h / 2
#choose center face
center_face_index = 0
min_dist = 9999999
gt_rect = ''
for j in range(face_count):
gt_rect = contents['faces'][j]['rect']
face_center_x = int(gt_rect[0]) + int(gt_rect[2] / 2)
face_center_y = int(gt_rect[1]) + int(gt_rect[3] / 2)
dist = math.sqrt(
math.pow(face_center_x - img_center_x, 2) +
math.pow(face_center_y - img_center_y, 2))
if dist < min_dist:
min_dist = dist
center_face_index = j
gt_rect = contents['faces'][center_face_index]['rect']
boxes, points = aligner.align_face(img, [gt_rect])
box = boxes[0]
pts = points[0]
spl = img_name.split('/')
base_name = spl[-1]
save_img_subdir = osp.join(save_aligned_dir, spl[-2])
if not osp.exists(save_img_subdir):
os.makedirs(save_img_subdir)
save_img_fn = osp.join(save_img_subdir, base_name)
print('save_img_fn: %s' % save_img_fn)
facial5points = np.reshape(pts, (2, -1))
dst_img = warp_and_crop_face(img, facial5points, reference_5pts,
output_size)
cv2.imwrite(save_img_fn, dst_img)
def main(nsplits,
split_id,
list_file,
img_root_dir,
mtcnn_model_dir,
save_dir=None,
rects_fn=None):
if not save_dir:
save_dir = './aligned_root_dir'
if not osp.exists(save_dir):
print('mkdir for aligned root dir: ', save_dir)
os.makedirs(save_dir)
save_aligned_dir = osp.join(save_dir, 'aligned_imgs')
if not osp.exists(save_aligned_dir):
print('mkdir for aligned/cropped face imgs: ', save_dir)
os.makedirs(save_aligned_dir)
save_rects_dir = osp.join(save_dir, 'face_rects')
if not osp.exists(save_rects_dir):
print('mkdir for face rects/landmarks: ', save_rects_dir)
os.makedirs(save_rects_dir)
aligner = MtcnnAligner(mtcnn_model_dir)
fp = open(list_file, 'r')
all_lines = fp.readlines()
fp.close()
rects_list = load_rect_list(rects_fn)
total_line_cnt = len(all_lines)
print('--->%d imgs in total' % total_line_cnt)
if nsplits < 2:
if split_id > 0:
print('===> Will only process first %d imgs' % split_id)
start_line = 0
end_line = split_id
else:
print('===> Will process all of the images')
start_line = 0
end_line = total_line_cnt
else:
assert (split_id < nsplits)
lines_per_split = float(total_line_cnt) / nsplits
start_line = int(lines_per_split * split_id)
end_line = int(lines_per_split * (split_id + 1))
if end_line + 1 >= total_line_cnt:
end_line = total_line_cnt
print('===> Will only process imgs in the range [%d, %d)]' %
(start_line, end_line))
count = start_line
for line in all_lines[start_line:end_line]:
line = line.strip()
print count
count = count + 1
img_fn = osp.join(img_root_dir, line)
print('===> Processing img: ' + img_fn)
img = cv2.imread(img_fn)
ht = img.shape[0]
wd = img.shape[1]
print 'image.shape:', img.shape
spl = osp.split(line)
sub_dir = osp.split(spl[0])[1]
print 'sub_dir: ', sub_dir
if CHINESE_2_PINYIN:
sub_dir = pinyin.get(sub_dir, format="strip")
# replace the dot sign in names
sub_dir = sub_dir.replace(u'\xb7', '-').encode('utf-8')
base_name = osp.splitext(spl[1])[0]
save_img_subdir = osp.join(save_aligned_dir, sub_dir)
if not osp.exists(save_img_subdir):
os.mkdir(save_img_subdir)
save_rect_subdir = osp.join(save_rects_dir, sub_dir)
if not osp.exists(save_rect_subdir):
os.mkdir(save_rect_subdir)
# print pts
save_rects_fn = osp.join(save_rect_subdir, base_name + '.txt')
fp_rect = open(save_rects_fn, 'w')
rect = get_rects_for_image(rects_list, base_name)
boxes, points = aligner.align_face(img, [rect])
nfaces = len(boxes)
fp_rect.write('%d\n' % nfaces)
for i in range(nfaces):
box = boxes[i]
pts = points[i]
if i:
save_img_fn = osp.join(save_img_subdir,
base_name + '_%d.jpg' % (i + 1))
else:
save_img_fn = osp.join(save_img_subdir, base_name + '.jpg')
facial5points = np.reshape(pts, (2, -1))
dst_img = warp_and_crop_face(img, facial5points, reference_5pts,
output_size)
cv2.imwrite(save_img_fn, dst_img)
print 'aligend face saved into: ', save_img_fn
for it in box:
fp_rect.write('%5.2f\t' % it)
fp_rect.write('\n')
for i in range(5):
fp_rect.write('%5.2f\t%5.2f\n' %
(facial5points[0][i], facial5points[1][i]))
fp_rect.close()
def main(args):
rect_root_dir = args.rect_root_dir
save_dir = args.save_dir
mtcnn_model_dir = args.mtcnn_model_dir
img_root_dir = args.image_root_dir
gpu_id = args.gpu_id
rect_list = get_rect_list(rect_root_dir)
if not save_dir:
save_dir = './aligned_root_dir'
if not osp.exists(save_dir):
print('makedirs for aligned root dir: ', save_dir)
os.makedirs(save_dir)
save_aligned_dir = osp.join(save_dir, 'aligned_imgs')
if not osp.exists(save_aligned_dir):
print('makedirs for aligned/cropped face imgs: ', save_dir)
os.makedirs(save_aligned_dir)
save_rects_dir = osp.join(save_dir, 'face_rects')
if not osp.exists(save_rects_dir):
print('makedirs for face rects/landmarks: ', save_rects_dir)
os.makedirs(save_rects_dir)
aligner = MtcnnAligner(mtcnn_model_dir, True, gpu_id=gpu_id)
for i in range(len(rect_list)):
with open(osp.join(rect_root_dir, rect_list[i]), 'r') as f:
lines = f.readlines()
for line in lines:
img_name = line.split(',')[0].replace('\\', '/')
gt_rect = get_gt_rect(line)
if gt_rect is None:
print('Failed to get_gt_rect(), skip to next image')
continue
img_fn = osp.join(img_root_dir, img_name)
print('===>Processing img: ' + img_fn)
img = cv2.imread(img_fn)
ht = img.shape[0]
wd = img.shape[1]
print 'image.shape:', img.shape
boxes, points = aligner.align_face(img, [gt_rect])
box = boxes[0]
pts = points[0]
spl = img_name.split('/')
sub_dir = ''
for j in range(len(spl) - 1):
sub_dir = osp.join(sub_dir, spl[j])
base_name = spl[-1]
save_img_subdir = osp.join(save_aligned_dir, sub_dir)
if not osp.exists(save_img_subdir):
os.makedirs(save_img_subdir)
save_img_fn = osp.join(save_img_subdir, base_name)
facial5points = np.reshape(pts, (2, -1))
dst_img = warp_and_crop_face(img, facial5points,
reference_5pts, output_size)
cv2.imwrite(save_img_fn, dst_img)
def main(face_json_file, mtcnn_model_dir, save_dir=None):
if not osp.exists(save_dir):
print('mkdir for aligned faces, aligned root dir: ', save_dir)
os.makedirs(save_dir)
if not osp.exists(aligned_save_dir):
print('mkdir for aligned faces, aligned root dir: ', save_dir)
os.makedirs(aligned_save_dir)
aligner = MtcnnAligner(mtcnn_model_dir, False)
fp = open('./identity_json_list_new.json', 'r')
all_lines = fp.readlines()
print len(all_lines)
count = 3432946
for line in all_lines[count:3600000]:
err_msg = ''
print count
#print line
count = count + 1
data = json.loads(line)
#print data
#origin_image_url = data[u'url']
image_url = data[u'url']
#splits = origin_image_url.split('/')
#image_url = splits[0] + '//' + osp.join(splits[2],splits[3],splits[4],splits[6],splits[7],splits[8])
#print '*******',image_url
gt = data[u'label'][u'detect'][u'general_d'][u'bbox'][0][u'pts']
print image_url
overlap_thresh_0 = overlap_thresh
save_path = osp.join(data[u'url'].split('/')[-3],
data[u'url'].split('/')[-2])
#print save_path
isExists1 = osp.exists(osp.join(save_dir, save_path))
if not isExists1:
os.makedirs(osp.join(save_dir, save_path))
save_img_fn = osp.join(data[u'url'].split('/')[-3],
data[u'url'].split('/')[-2])
save_fn = aligned_save_dir + '/' + save_img_fn + '/' + data[
u'url'].split('/')[-1]
#print save_fn
isExists2 = osp.exists(osp.join(aligned_save_dir, save_img_fn))
if not isExists2:
os.makedirs(osp.join(aligned_save_dir, save_img_fn))
filename = data[u'url'].split('/')[-1][:-4] + '.json'
print filename
fp_rlt = open(osp.join(save_dir, save_path, filename), 'w')
item = {}
item[u'url'] = image_url
try:
#print('===> Processing image: ' + save_path + '/' + filename)
imageBGR = io.imread(image_url)
image = cv2.cvtColor(imageBGR, cv2.COLOR_BGR2RGB)
print image.shape
#GT_RECT = [0,0,image.shape[0],image.shape[1]]
#GT_RECT = [item['gt'][0][0],item['gt'][0][1],item['gt'][2][0],item['gt'][2][1]]
GT_RECT = [gt[0][0], gt[0][1], gt[2][0], gt[2][1]]
#print 'gt',gt
#print 'GT_RECT',GT_RECT
boxes, points = aligner.align_face(image, [GT_RECT])
box = boxes[0]
pts = points[0]
#print pts
facial5points = np.reshape(points, (2, -1))
dst_img = warp_and_crop_face(image, facial5points, reference_5pts,
output_size)
cv2.imwrite(save_fn, dst_img)
#print '########'
print('===> Processing image: ' + save_path + '/' + filename)
tmp = {'rect': box[0:4], 'score': box[4], 'gt': gt, 'pts': pts}
item['faces'] = tmp
item['id'] = data[u'url'].split('/')[-2]
item['shape'] = image.shape
except:
continue
json.dump(item, fp_rlt, indent=4)
fp_rlt.close()
def main(face_json_file, img_root_dir, mtcnn_model_dir, save_dir=None):
if save_dir is None:
save_dir = './fd_json_add_missed'
fp_in = open(face_json_file, 'r')
img_list = json.load(fp_in)
fp_in.close()
if not osp.exists(img_root_dir):
print('ERROR: Cannot find image root dir: ' + img_root_dir)
return
if not osp.exists(save_dir):
print('mkdir for aligned faces, aligned root dir: ', save_dir)
os.makedirs(save_dir)
json_basename = osp.basename(face_json_file)
splits = osp.splitext(json_basename)
new_json_fn = splits[0] + '_add_missed' + splits[1]
aligner = MtcnnAligner(mtcnn_model_dir, False)
fp_log1 = open(osp.join(save_dir, log_fn1), 'w')
fp_log2 = open(osp.join(save_dir, log_fn2), 'w')
fp_rlt = open(osp.join(save_dir, new_json_fn), 'w')
missed_count1 = 0
missed_count2 = 0
for item in img_list:
err_msg = ''
if 'filename' not in item:
err_msg = "'filename' not in item, break..."
print(err_msg)
fp_log2.write(err_msg + '\n')
break
img_fn = osp.join(img_root_dir, item['filename'])
save_fn = osp.join(save_dir, item['filename'])
save_fn_dir = osp.dirname(save_fn)
overlap_thresh_0 = overlap_thresh
## Tom_Brady_0002 is special cauz the face in the image is very small
if 'Tom_Brady_0002' in img_fn:
overlap_thresh_0 = 0.25
print('===> Processing image: ' + img_fn)
if 'faces' not in item:
err_msg = "'faces' not in item"
fp_log2.write(item['filename'] + ': ' + err_msg + '\n')
continue
elif 'face_count' not in item:
err_msg = "'face_count' not in item"
fp_log2.write(item['filename'] + ': ' + err_msg + '\n')
continue
nfaces = item['face_count']
if nfaces < 1:
missed_count1 += 1
fp_log2.write(item['filename'] + ': ' + "item['face_count'] < 1" +
'\n')
img = cv2.imread(img_fn)
boxes, points = aligner.align_face(img, [GT_RECT])
item['face_count'] = 1
box = boxes[0]
pts = points[0]
tmp = {'rect': box[0:4], 'score': box[4], 'pts': pts}
item['faces'].append(tmp)
item['message'] = 'success'
item['used_gt'] = 1
fp_log2.write('-->faces added by aligner:\n{}\n'.format(tmp))
continue
overlaps = get_gt_overlap(item['faces'])
max_overlap_idx = overlaps.argmax()
if overlaps[max_overlap_idx] >= overlap_thresh_0:
fp_log1.write(item['filename'] + ': ' + " max_overlap_idx=" +
str(max_overlap_idx) + '\n')
else:
missed_count2 += 1
fp_log2.write(item['filename'] + ': ' +
"no faces have overlap>={} with groundtruth".format(
overlap_thresh_0) + '\n')
fp_log2.write("--> max_overlap_idx = {}\n".format(max_overlap_idx))
fp_log2.write("--> overlaps = {}\n".format(overlaps))
img = cv2.imread(img_fn)
boxes, points = aligner.align_face(img, [GT_RECT])
item['face_count'] += 1
box = boxes[0]
pts = points[0]
tmp = {'rect': box[0:4], 'score': box[4], 'pts': pts}
item['faces'].append(tmp)
item['used_gt'] = 1
fp_log2.write('-->faces added by aligner:\n{}\n'.format(tmp))
fp_log2.write("\n==>Images with missed faces: {}\n".format(missed_count1 +
missed_count2))
fp_log2.write(
"\t{} missed because of no detection\n".format(missed_count1))
fp_log2.write(
"\t{} missed because of max_overlap<thresh\n".format(missed_count2))
json.dump(img_list, fp_rlt, indent=2)
fp_rlt.close()
fp_log1.close()
fp_log2.close()
def __init__(self, caffe_model_path):
self.aligner = MtcnnAligner(caffe_model_path)
def main(img_list_file, root_dir, mtcnn_model_dir, save_dir=None):
if not save_dir:
save_dir = './aligned_images'
if not osp.exists(save_dir):
print('mkdir for aligned faces, aligned root dir: ', save_dir)
os.makedirs(save_dir)
aligned_save_dir = osp.join(save_dir, 'aligned_faces')
if not osp.exists(aligned_save_dir):
print('mkdir for aligned faces, aligned images dir: ', aligned_save_dir)
os.makedirs(aligned_save_dir)
aligner = MtcnnAligner(mtcnn_model_dir, False)
fp = open(img_list_file,'r')
fn_rlt = osp.join(save_dir, 'fd_rlt.json')
fp_rlt = open(fn_rlt, 'w')
fp_rlt.write('[\n')
count = 0
for line in fp:
print line
line_split = line.split()
img_fn = line_split[0]
id_num = line_split[1]
img_fn_split = img_fn.split('/')
img_full_fn = osp.join(root_dir, img_fn)
print 'process image: ', img_full_fn, " id_num: ", id_num
#for root,dirs,files in path_walk:
err_msg = ''
if not count:
fp_rlt.write(',\n')
count = count + 1
print 'count: ', count
overlap_thresh_0 = overlap_thresh
save_subdir = osp.join(aligned_save_dir, img_fn_split[-2])
save_img_fn = osp.join(save_subdir, img_fn_split[-1])
if not osp.exists(save_subdir):
os.makedirs(save_subdir)
image = cv2.imread(img_full_fn)
print image.shape
boxes, points = aligner.align_face(image, [GT_RECT])
box = boxes[0]
pts = points[0]
facial5points = np.reshape(points, (2, -1))
dst_img = warp_and_crop_face(image, facial5points, reference_5pts, output_size)
cv2.imwrite(save_img_fn, dst_img)
item = {
'filename': img_base_fn,
'face_count': 1,
}
tmp = {'rect': box[0:4],
'score': box[4],
'pts': pts,
'id': id_num
}
item['faces'] = [tmp]
#item['id'] = data[u'url'].splitit('/')[-3]
item['shape'] = image.shape
json_str = json.dumps(item, indent=2)
fp_rlt.write(json_str+'\n')
fp_rlt.flush()
fp_rlt.write(']\n')
fp_rlt.close()
fp.close()
def main(args):
rect_root_dir = args.rect_root_dir
save_dir = args.save_dir
mtcnn_model_dir = args.mtcnn_model_dir
img_root_dir = args.image_root_dir
gpu_id = args.gpu_id
rect_list = get_rect_list(rect_root_dir)
print('all %d images' % (len(rect_list)))
if not save_dir:
save_dir = './aligned_root_dir'
if not osp.exists(save_dir):
print('makedirs for aligned root dir: ', save_dir)
os.makedirs(save_dir)
save_aligned_dir = osp.join(save_dir, 'aligned_imgs')
if not osp.exists(save_aligned_dir):
print('makedirs for aligned/cropped face imgs: ', save_dir)
os.makedirs(save_aligned_dir)
save_rects_dir = osp.join(save_dir, 'face_rects')
if not osp.exists(save_rects_dir):
print('makedirs for face rects/landmarks: ', save_rects_dir)
os.makedirs(save_rects_dir)
aligner = MtcnnAligner(mtcnn_model_dir, True, gpu_id=gpu_id)
for i in range(len(rect_list)):
with open(osp.join(rect_root_dir, rect_list[i]), 'r') as f:
#lines = f.readlines()
contents = json.load(f)
img_name = contents['filename']
gt_rect = contents['faces'][0]['rect']
if gt_rect is None:
print('Failed to get_gt_rect(), skip to next image')
continue
img = cv2.imread(img_name)
ht = img.shape[0]
wd = img.shape[1]
print 'image.shape:', img.shape
boxes, points = aligner.align_face(img, [gt_rect])
box = boxes[0]
pts = points[0]
spl = img_name.split('/')
base_name = spl[-1]
save_img_subdir = osp.join(save_aligned_dir, spl[-2])
if not osp.exists(save_img_subdir):
os.makedirs(save_img_subdir)
save_img_fn = osp.join(save_img_subdir, base_name)
print('save_img_fn: %s' % save_img_fn)
facial5points = np.reshape(pts, (2, -1))
dst_img = warp_and_crop_face(img, facial5points, reference_5pts,
output_size)
cv2.imwrite(save_img_fn, dst_img)
def main(args):
save_dir = args.save_dir
list_file = args.image_list
nsplits = args.nsplits
split_id = args.split_id
mtcnn_model_dir = args.mtcnn_model_dir
img_root_dir = args.image_root_dir
rect_root_dir = args.rect_root_dir
gpu_id = args.gpu_id
if not save_dir:
save_dir = './aligned_root_dir'
if not osp.exists(save_dir):
print('makedirs for aligned root dir: ', save_dir)
os.makedirs(save_dir)
save_aligned_dir = osp.join(save_dir, 'aligned_imgs')
if not osp.exists(save_aligned_dir):
print('makedirs for aligned/cropped face imgs: ', save_dir)
os.makedirs(save_aligned_dir)
save_rects_dir = osp.join(save_dir, 'face_rects')
if not osp.exists(save_rects_dir):
print('makedirs for face rects/landmarks: ', save_rects_dir)
os.makedirs(save_rects_dir)
aligner = MtcnnAligner(mtcnn_model_dir, True, gpu_id=gpu_id)
fp = open(list_file, 'r')
all_lines = fp.readlines()
fp.close()
total_line_cnt = len(all_lines)
print('--->%d imgs in total' % total_line_cnt)
if nsplits < 2:
if split_id > 0:
print('===> Will only process first %d imgs' % split_id)
start_line = 0
end_line = split_id
else:
print('===> Will process all of the images')
start_line = 0
end_line = total_line_cnt
else:
assert (split_id < nsplits)
lines_per_split = float(total_line_cnt) / nsplits
start_line = int(lines_per_split * split_id)
end_line = int(lines_per_split * (split_id + 1))
if end_line + 1 >= total_line_cnt:
end_line = total_line_cnt
print('===> Will only process imgs in the range [%d, %d)]' %
(start_line, end_line))
count = start_line
all_lines = all_lines[start_line:end_line]
for line in all_lines:
line = line.strip()
print '%d\n' % count
count = count + 1
# img_fn = osp.join(img_root_dir, line)
img_fn = line
print('===> Processing img: ' + img_fn)
img = cv2.imread(img_fn)
ht = img.shape[0]
wd = img.shape[1]
print 'image.shape:', img.shape
rect_fn = osp.join(
rect_root_dir,
rename_with_folderName(line) +
'.json') #osp.join(rect_root_dir,line[0:-4] + '.json')
gt_rect = get_gt_rect(rect_fn)
if gt_rect is None:
print('Failed to get_gt_rect(), skip to next image')
continue
# print 'face rect: ', gt
boxes, points = aligner.align_face(img, [gt_rect])
assert len(boxes) == len(points)
for i in range(len(boxes)):
box = boxes[i]
pts = points[i]
spl = osp.split(line)
sub_dir = spl[0]
base_name = spl[-1]
save_img_subdir = osp.join(save_aligned_dir, sub_dir)
if not osp.exists(save_img_subdir):
os.makedirs(save_img_subdir)
save_rect_subdir = osp.join(save_rects_dir, sub_dir)
if not osp.exists(save_rect_subdir):
os.makedirs(save_rect_subdir)
# print pts
save_img_fn = osp.join(save_img_subdir, base_name)
facial5points = np.reshape(pts, (2, -1))
dst_img = warp_and_crop_face(img, facial5points, reference_5pts,
output_size)
cv2.imwrite(save_img_fn, dst_img)
save_rect_fn = osp.join(save_rect_subdir,
osp.splitext(base_name)[0] + '.txt')
fp_rect = open(save_rect_fn, 'w')
for it in box:
fp_rect.write('%5.2f\t' % it)
fp_rect.write('\n')
for i in range(5):
fp_rect.write('%5.2f\t%5.2f\n' %
(facial5points[0][i], facial5points[1][i]))
fp_rect.close()
def main(nsplits, split_id, list_file, img_root_dir, mtcnn_model_dir, save_dir=None):
if not save_dir:
save_dir = './aligned_root_dir'
if not osp.exists(save_dir):
print('makedirs for aligned root dir: ', save_dir)
os.makedirs(save_dir)
save_aligned_dir = osp.join(save_dir, 'aligned_imgs')
if not osp.exists(save_aligned_dir):
print('makedirs for aligned/cropped face imgs: ', save_dir)
os.makedirs(save_aligned_dir)
save_rects_dir = osp.join(save_dir, 'face_rects')
if not osp.exists(save_rects_dir):
print('makedirs for face rects/landmarks: ', save_rects_dir)
os.makedirs(save_rects_dir)
aligner = MtcnnAligner(mtcnn_model_dir, False)
fp = open(list_file, 'r')
all_lines = fp.readlines()
fp.close()
total_line_cnt = len(all_lines)
print('--->%d imgs in total' % total_line_cnt)
if nsplits < 2:
if split_id > 0:
print('===> Will only process first %d imgs' % split_id)
start_line = 0
end_line = split_id
else:
print('===> Will process all of the images')
start_line = 0
end_line = total_line_cnt
else:
assert(split_id < nsplits)
lines_per_split = float(total_line_cnt) / nsplits
start_line = int(lines_per_split * split_id)
end_line = int(lines_per_split * (split_id + 1))
if end_line + 1 >= total_line_cnt:
end_line = total_line_cnt
print('===> Will only process imgs in the range [%d, %d)]' % (
start_line, end_line))
count = start_line
for line in all_lines[start_line:end_line]:
line = line.strip()
print count
count = count + 1
img_fn = osp.join(img_root_dir, line)
print('===> Processing img: ' + img_fn)
img = cv2.imread(img_fn)
ht = img.shape[0]
wd = img.shape[1]
print 'image.shape:', img.shape
# GT_RECT = [0,0,img.shape[0],img.shape[1]]
# GT_RECT = [int(wd * 0.25), int(ht * 0.25),
# int(wd * 0.75), int(ht * 0.72)]
GT_RECT = get_gt_rect(img)
# print 'face rect: ', gt
boxes, points = aligner.align_face(img, [GT_RECT])
box = boxes[0]
pts = points[0]
spl = osp.split(line)
sub_dir = spl[0]
base_name = spl[-1]
save_img_subdir = osp.join(save_aligned_dir, sub_dir)
if not osp.exists(save_img_subdir):
os.makedirs(save_img_subdir)
save_rect_subdir = osp.join(save_rects_dir, sub_dir)
if not osp.exists(save_rect_subdir):
os.makedirs(save_rect_subdir)
# print pts
save_img_fn = osp.join(save_img_subdir, base_name)
facial5points = np.reshape(pts, (2, -1))
dst_img = warp_and_crop_face(
img, facial5points, reference_5pts, output_size)
cv2.imwrite(save_img_fn, dst_img)
save_rect_fn = osp.join(
save_rect_subdir, osp.splitext(base_name)[0] + '.txt')
fp_rect = open(save_rect_fn, 'w')
for it in box:
fp_rect.write('%5.2f\t' % it)
fp_rect.write('\n')
for i in range(5):
fp_rect.write('%5.2f\t%5.2f\n' %
(facial5points[0][i], facial5points[1][i]))
fp_rect.close()
def __init__(self, caffe_model_path=None, gpu_id=0):
self.aligner = None
if caffe_model_path:
self.aligner = MtcnnAligner(caffe_model_path, gpu_id)
def main(json_file, save_dir=None, save_img=True, show_img=True):
if not osp.exists(json_file):
print 'Cannot find json file: ' + json_file
pass
if save_dir is None:
save_dir = './fa_facex_rlt'
save_json = 'mtcnn_align_rlt.json'
caffe_model_path = "../../model"
fp_json = open(json_file, 'r')
facex_response = json.load(fp_json)
fp_json.close()
if (not facex_response or not isinstance(facex_response, dict)
or 'facex_det' not in facex_response):
print 'Invalid json file: ' + json_file
pass
facex_det_response = facex_response['facex_det']
if not osp.exists(save_dir):
os.makedirs(save_dir)
fp_rlt = open(osp.join(save_dir, save_json), 'w')
results = []
for item in facex_det_response:
img_path = item['name']
print '===> Processing image: ' + img_path
if 'detections' not in item:
continue
face_rects = []
for face in item['detections']:
face_rects.append(face['pts'])
img = cv2.imread(img_path)
aligner = MtcnnAligner(caffe_model_path, False)
rlt = {}
rlt["filename"] = img_path
rlt["faces"] = []
rlt['face_count'] = 0
t1 = time.clock()
bboxes, points = aligner.align_face(img, face_rects)
t2 = time.clock()
n_boxes = len(face_rects)
print(
"-->Alignment cost %f seconds, processed %d face rects, avg time: %f seconds"
% ((t2 - t1), n_boxes, (t2 - t1) / n_boxes))
if bboxes is not None and len(bboxes) > 0:
for (box, pts) in zip(bboxes, points):
# box = box.tolist()
# pts = pts.tolist()
tmp = {'rect': box[0:4], 'score': box[4], 'pts': pts}
rlt['faces'].append(tmp)
rlt['face_count'] = len(bboxes)
rlt['message'] = 'success'
results.append(rlt)
if save_img or show_img:
draw_faces(img, bboxes, points)
if save_img:
pdir = save_dir
dir_name, base_name = osp.split(img_path)
if dir_name:
_, pdir = osp.split(dir_name)
if pdir:
pdir = osp.join(save_dir, pdir)
if not osp.exists(pdir):
os.makedirs(pdir)
save_name = osp.join(pdir, osp.basename(img_path))
cv2.imwrite(save_name, img)
if show_img:
cv2.imshow('img', img)
cv2.waitKey(0)
cv2.destroyAllWindows()
json.dump(results, fp_rlt, indent=4)
fp_rlt.close()
def __init__(self, model_path=None, gpu_id=0):
self.aligner = None
if model_path:
self.aligner = MtcnnAligner(model_path, gpu_id=gpu_id)
def main(mtcnn_model_dir, save_dir=None, save_img=None):
if save_dir is None:
save_dir = './gender_origin/label'
if not osp.exists(save_dir):
os.makedirs(save_dir)
if save_img is None:
save_img = './gender_origin/img'
if not osp.exists(save_img):
os.makedirs(save_img)
aligner = MtcnnAligner(mtcnn_model_dir, False)
fp = open(
'/disk2/du/mtcnn-caffe-good/mtcnn_aligner/dataset-gender-all.json',
'r')
all_lines = fp.readlines()
count = 23097
for line in all_lines[23097:]:
count = count + 1
print count
data = json.loads(line)
image_url = data["url"]
image_bbox = data["label"]["detect"]["general_d"]["bbox"]
file_idx = "%06d" % count
print file_idx
filename = str(file_idx) + '.json'
imgname = str(file_idx) + '.jpg'
urllib.urlretrieve(image_url, osp.join(save_img, imgname))
print('===> Save image: ' + filename)
fp_rlt = open(osp.join(save_dir, filename), 'w')
item = {}
item['url'] = image_url
item['imgname'] = imgname
if image_bbox == []:
item['detect'] = []
#continue
else:
img = io.imread(image_url)
if len(img.shape) != 3:
item['detect'] = []
#continue
else:
item['detect'] = []
for idx in range(len(image_bbox)):
print "len of box", len(image_bbox)
#item['detect'] = []
per_bbox = image_bbox[idx]["pts"]
GT_RECT = [
per_bbox[0][0], per_bbox[0][1], per_bbox[2][0],
per_bbox[2][1]
]
boxes, points = aligner.align_face(img, [GT_RECT])
pts = points[0]
label = image_bbox[idx]["class"]
tmp = {'rect': GT_RECT, 'pts': pts, 'class': label}
item['detect'].append(tmp)
print "item:", item
json.dump(item, fp_rlt, indent=4)
fp_rlt.close()