def main():
# get updated faces properties and grade weight
groups = extractFacesProperties()
face_attrs, img_attrs = getUpdatedGradesWeight()
best_faces = []
for group in groups:
for face in group:
face.set_grade(face_attrs)
best_faces.append(max(group, key = lambda face: face.grade).face_info)
count_bet_faces_in_image = {}
for best_face in best_faces:
curr = count_bet_faces_in_image.get(best_face["image_index"], 0)
count_bet_faces_in_image[best_face["image_index"]] = curr + 1
print(count_bet_faces_in_image)
# get images with the highest number of good faces
best_image_index = [k for k,v in count_bet_faces_in_image.items() if v == max(count_bet_faces_in_image.values())]
# if there are more than 1 image with highest number of good faces
if len(best_image_index) > 1:
with open(FGFI_DIR + '/output/data.json', 'r') as f:
image_data_json = json.load(f)
best_images_props = []
best_images = []
for best_image in best_image_index:
best_images_props.append(set_image_properties(image_data_json["images"][best_image]["path"]))
best_images.append(imageToClass(best_images_props[best_image], best_image))
for image_prop in best_images:
image_prop.set_grade(img_attrs)
best_image_index = max(best_images, key=operator.attrgetter('grade')).index
else:
best_image_index = max(count_bet_faces_in_image.items(), key = operator.itemgetter(1))[0]
faces_to_swap = []
for best_face in best_faces:
if best_face["image_index"] != best_image_index:
faces_to_swap.append({
"group_index": best_face["group_index"],
"group_face_index": best_face["group_face_index"]
})
utils.save_to_json_file(os.path.join(OUTPUT_DIR, "input.json"),
json.dumps( {
"image_index": best_image_index,
"faces": faces_to_swap
}))
def detect_properties(images):
"""Detects image properties in the file."""
from google.cloud import vision
client = vision.ImageAnnotatorClient()
for pic in images:
with io.open(pic, 'rb') as image_file:
content = image_file.read()
image = vision.types.Image(content=content)
response = client.image_properties(image=image)
utils.save_to_json_file(
os.path.join(OUTPUT_DIR,
pic.split("/input/")[1].split(".")[0] + ".json"),
MessageToJson(response))
def main():
images = get_images()
images_info = crop_faces_from_images(images)
groups_dict = get_cropped_images_groups(images_info)
#Convert groups to list of dictionary
groups = [[] for i in range(len(set(groups_dict.values())))]
for image_face_index, group_index in groups_dict.items():
image_index, face_index = image_face_index.split("_")
groups[group_index].append({
"image_index": int(image_index),
"face_index": int(face_index)
})
#Save the data to file
utils.save_to_json_file(
os.path.join(OUTPUT_DIR, "data.json"),
json.dumps({
"images": images_info,
"groups": groups
}))
os.makedirs(NOBG_DIR)
def removebg(image):
response = requests.post(
'https://api.remove.bg/v1.0/removebg',
files={'image_file': utils.image_to_bytes(image)},
data={'size': 'regular'},
headers={'X-Api-Key': API_KEY},
)
if response.status_code == requests.codes.ok:
return Image.open(io.BytesIO(response.content))
else:
print("Error:", response.status_code, response.text)
if __name__ == "__main__":
data = utils.from_json_file(os.path.join(JSON_DIR, "data.json"))
input = utils.from_json_file(os.path.join(DEBUG_DIR, "input.json"))
ngbg = []
for face in input["faces"]:
cropped = utils.get_cropped_image(data, face, config.FGFI_DIR)
ngbg.append({
"info":
face,
"image_data":
base64.b64encode(removebg(cropped, face)).decode("utf-8")
})
utils.save_to_json_file(os.path.join(OUTPUT_DIR, "nobg.json"),
json.dumps(ngbg))