def run_on_photo(photo_id):
model = ObjectModel()
sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
from photonix.classifiers.runners import results_for_model_on_photo, get_or_create_tag
photo, results = results_for_model_on_photo(model, photo_id)
if photo:
from django.utils import timezone
from photonix.photos.models import PhotoTag
photo.clear_tags(source='C', type='O')
for result in results:
tag = get_or_create_tag(library=photo.library,
name=result['label'],
type='O',
source='C')
PhotoTag(photo=photo,
tag=tag,
source='C',
confidence=result['score'],
significance=result['significance'],
position_x=result['x'],
position_y=result['y'],
size_x=result['width'],
size_y=result['height']).save()
photo.classifier_object_completed_at = timezone.now()
photo.classifier_object_version = getattr(model, 'version', 0)
photo.save()
return photo, results
def run_on_photo(photo_id):
model = StyleModel()
sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
from photonix.classifiers.runners import results_for_model_on_photo, get_or_create_tag
photo, results = results_for_model_on_photo(model, photo_id)
if photo:
from django.utils import timezone
from photonix.photos.models import PhotoTag
photo.clear_tags(source='C', type='S')
for name, score in results:
tag = get_or_create_tag(library=photo.library,
name=name,
type='S',
source='C')
PhotoTag(photo=photo,
tag=tag,
source='C',
confidence=score,
significance=score).save()
photo.classifier_style_completed_at = timezone.now()
photo.classifier_style_version = getattr(model, 'version', 0)
photo.save()
return photo, results
def run_on_photo(photo_id):
model = LocationModel()
sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
from photonix.classifiers.runners import results_for_model_on_photo, get_or_create_tag
photo, results = results_for_model_on_photo(model, photo_id)
if photo and results['country']:
from photonix.photos.models import PhotoTag
photo.clear_tags(source='C', type='L')
country_tag = get_or_create_tag(library=photo.library,
name=results['country']['name'],
type='L',
source='C')
PhotoTag(photo=photo,
tag=country_tag,
source='C',
confidence=1.0,
significance=1.0).save()
if results['city']:
city_tag = get_or_create_tag(library=photo.library,
name=results['city']['name'],
type='L',
source='C',
parent=country_tag)
PhotoTag(photo=photo,
tag=city_tag,
source='C',
confidence=0.5,
significance=0.5).save()
return photo, results
def run_on_photo(photo_id):
model = StyleModel()
sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
from photonix.classifiers.runners import results_for_model_on_photo, get_or_create_tag
photo, results = results_for_model_on_photo(model, photo_id)
if photo and results is not None:
from photonix.photos.models import PhotoTag
photo.clear_tags(source='C', type='S')
for name, score in results:
tag = get_or_create_tag(library=photo.library, name=name, type='S', source='C')
PhotoTag(photo=photo, tag=tag, source='C', confidence=score, significance=score).save()
return photo, results
def run_on_photo(photo_id):
sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
from photonix.classifiers.runners import get_photo_by_any_type, results_for_model_on_photo, get_or_create_tag
photo = get_photo_by_any_type(photo_id)
model = FaceModel(library_id=photo and photo.library_id)
# Detect all faces in an image
photo, results = results_for_model_on_photo(model, photo_id)
# Read image data so we can extract faces and create embeddings
path = photo_id
if photo:
path = photo.base_image_path
model.library_id = photo.library_id
image_data = Image.open(path)
# Loop over each face that was detected above
for result in results:
# Crop individual face + 30% extra in each direction
box = result['box']
face_image = model.crop(image_data, box)
# Generate embedding with Facenet
try:
embedding = model.get_face_embedding(face_image)
# Add it to the results
result['embedding'] = embedding
if photo:
closest_tag, closest_distance = model.find_closest_face_tag(
embedding)
if closest_tag:
print(f'Closest tag: {closest_tag}')
print(f'Closest distance: {closest_distance}')
result['closest_tag'] = closest_tag
result['closest_distance'] = closest_distance
except ValueError:
pass
if photo:
from django.utils import timezone
from photonix.photos.models import Tag, PhotoTag
photo.clear_tags(source='C', type='F')
for result in results:
# Use matched tag if within distance threshold
if result.get('closest_distance', 999) < DISTANCE_THRESHOLD:
tag = Tag.objects.get(id=result['closest_tag'],
library=photo.library,
type='F')
print(f'MATCHED {tag.name}')
# Otherwise create new tag
else:
while True:
random_name = f'Unknown person {randint(0, 999999):06d}'
try:
Tag.objects.get(library=photo.library,
name=random_name,
type='F',
source='C')
except Tag.DoesNotExist:
tag = Tag(library=photo.library,
name=random_name,
type='F',
source='C')
tag.save()
break
x = (result['box'][0] +
(result['box'][2] / 2)) / photo.base_file.width
y = (result['box'][1] +
(result['box'][3] / 2)) / photo.base_file.height
width = result['box'][2] / photo.base_file.width
height = result['box'][3] / photo.base_file.height
score = result['confidence']
extra_data = ''
if 'embedding' in result:
extra_data = json.dumps(
{'facenet_embedding': result['embedding']})
PhotoTag(photo=photo,
tag=tag,
source='F',
confidence=score,
significance=score,
position_x=x,
position_y=y,
size_x=width,
size_y=height,
model_version=model.version,
retrained_model_version=model.retrained_version,
extra_data=extra_data).save()
photo.classifier_color_completed_at = timezone.now()
photo.classifier_color_version = getattr(model, 'version', 0)
photo.save()
print('Finished')
return photo, results
