blakeblackshear.frigate/frigate/embeddings/maintainer.py

771 lines
28 KiB
Python

"""Maintain embeddings in SQLite-vec."""
import base64
import logging
import os
import threading
from multiprocessing.synchronize import Event as MpEvent
from typing import Optional
import cv2
import numpy as np
import requests
from peewee import DoesNotExist
from playhouse.sqliteq import SqliteQueueDatabase
from frigate.comms.embeddings_updater import EmbeddingsRequestEnum, EmbeddingsResponder
from frigate.comms.event_metadata_updater import (
EventMetadataSubscriber,
EventMetadataTypeEnum,
)
from frigate.comms.events_updater import EventEndSubscriber, EventUpdateSubscriber
from frigate.comms.inter_process import InterProcessRequestor
from frigate.config import FrigateConfig
from frigate.const import CLIPS_DIR, FRIGATE_LOCALHOST, UPDATE_EVENT_DESCRIPTION
from frigate.embeddings.alpr.alpr import LicensePlateRecognition
from frigate.events.types import EventTypeEnum
from frigate.genai import get_genai_client
from frigate.models import Event
from frigate.util.builtin import serialize
from frigate.util.image import SharedMemoryFrameManager, area, calculate_region
from .embeddings import Embeddings
logger = logging.getLogger(__name__)
REQUIRED_FACES = 2
MAX_THUMBNAILS = 10
class EmbeddingMaintainer(threading.Thread):
"""Handle embedding queue and post event updates."""
def __init__(
self,
db: SqliteQueueDatabase,
config: FrigateConfig,
stop_event: MpEvent,
) -> None:
super().__init__(name="embeddings_maintainer")
self.config = config
self.embeddings = Embeddings(config, db)
# Check if we need to re-index events
if config.semantic_search.reindex:
self.embeddings.reindex()
self.event_subscriber = EventUpdateSubscriber()
self.event_end_subscriber = EventEndSubscriber()
self.event_metadata_subscriber = EventMetadataSubscriber(
EventMetadataTypeEnum.regenerate_description
)
self.embeddings_responder = EmbeddingsResponder()
self.frame_manager = SharedMemoryFrameManager()
# set face recognition conditions
self.face_recognition_enabled = self.config.face_recognition.enabled
self.requires_face_detection = "face" not in self.config.objects.all_objects
self.detected_faces: dict[str, float] = {}
# create communication for updating event descriptions
self.requestor = InterProcessRequestor()
self.stop_event = stop_event
self.tracked_events: dict[str, list[any]] = {}
self.genai_client = get_genai_client(config)
# set license plate recognition conditions
self.lpr_config = self.config.lpr
self.requires_license_plate_detection = (
"license_plate" not in self.config.objects.all_objects
)
self.detected_license_plates: dict[str, dict[str, any]] = {}
if self.lpr_config.enabled:
self.license_plate_recognition = LicensePlateRecognition(
self.lpr_config, self.requestor, self.embeddings
)
@property
def face_detector(self) -> cv2.FaceDetectorYN:
# Lazily create the classifier.
if "face_detector" not in self.__dict__:
self.__dict__["face_detector"] = cv2.FaceDetectorYN.create(
"/config/model_cache/facenet/facedet.onnx",
config="",
input_size=(320, 320),
score_threshold=0.8,
nms_threshold=0.3,
)
return self.__dict__["face_detector"]
def run(self) -> None:
"""Maintain a SQLite-vec database for semantic search."""
while not self.stop_event.is_set():
self._process_requests()
self._process_updates()
self._process_finalized()
self._process_event_metadata()
self.event_subscriber.stop()
self.event_end_subscriber.stop()
self.event_metadata_subscriber.stop()
self.embeddings_responder.stop()
self.requestor.stop()
logger.info("Exiting embeddings maintenance...")
def _process_requests(self) -> None:
"""Process embeddings requests"""
def _handle_request(topic: str, data: dict[str, any]) -> str:
try:
if topic == EmbeddingsRequestEnum.embed_description.value:
return serialize(
self.embeddings.embed_description(
data["id"], data["description"]
),
pack=False,
)
elif topic == EmbeddingsRequestEnum.embed_thumbnail.value:
thumbnail = base64.b64decode(data["thumbnail"])
return serialize(
self.embeddings.embed_thumbnail(data["id"], thumbnail),
pack=False,
)
elif topic == EmbeddingsRequestEnum.generate_search.value:
return serialize(
self.embeddings.text_embedding([data])[0], pack=False
)
elif topic == EmbeddingsRequestEnum.register_face.value:
if data.get("cropped"):
self.embeddings.embed_face(
data["face_name"],
base64.b64decode(data["image"]),
upsert=True,
)
return True
else:
img = cv2.imdecode(
np.frombuffer(
base64.b64decode(data["image"]), dtype=np.uint8
),
cv2.IMREAD_COLOR,
)
face_box = self._detect_face(img)
if not face_box:
return False
face = img[face_box[1] : face_box[3], face_box[0] : face_box[2]]
ret, webp = cv2.imencode(
".webp", face, [int(cv2.IMWRITE_WEBP_QUALITY), 100]
)
self.embeddings.embed_face(
data["face_name"], webp.tobytes(), upsert=True
)
return False
except Exception as e:
logger.error(f"Unable to handle embeddings request {e}")
self.embeddings_responder.check_for_request(_handle_request)
def _process_updates(self) -> None:
"""Process event updates"""
update = self.event_subscriber.check_for_update(timeout=0.01)
if update is None:
return
source_type, _, camera, data = update
if not camera or source_type != EventTypeEnum.tracked_object:
return
camera_config = self.config.cameras[camera]
# no need to process updated objects if face recognition, lpr, genai are disabled
if (
not camera_config.genai.enabled
and not self.face_recognition_enabled
and not self.lpr_config.enabled
):
return
# Create our own thumbnail based on the bounding box and the frame time
try:
frame_id = f"{camera}{data['frame_time']}"
yuv_frame = self.frame_manager.get(frame_id, camera_config.frame_shape_yuv)
except FileNotFoundError:
pass
if yuv_frame is None:
logger.debug(
"Unable to process object update because frame is unavailable."
)
return
if self.face_recognition_enabled:
self._process_face(data, yuv_frame)
if self.lpr_config.enabled:
self._process_license_plate(data, yuv_frame)
# no need to save our own thumbnails if genai is not enabled
# or if the object has become stationary
if self.genai_client is not None and not data["stationary"]:
if data["id"] not in self.tracked_events:
self.tracked_events[data["id"]] = []
data["thumbnail"] = self._create_thumbnail(yuv_frame, data["box"])
# Limit the number of thumbnails saved
if len(self.tracked_events[data["id"]]) >= MAX_THUMBNAILS:
# Always keep the first thumbnail for the event
self.tracked_events[data["id"]].pop(1)
self.tracked_events[data["id"]].append(data)
self.frame_manager.close(frame_id)
def _process_finalized(self) -> None:
"""Process the end of an event."""
while True:
ended = self.event_end_subscriber.check_for_update(timeout=0.01)
if ended == None:
break
event_id, camera, updated_db = ended
camera_config = self.config.cameras[camera]
if event_id in self.detected_faces:
self.detected_faces.pop(event_id)
if event_id in self.detected_license_plates:
self.detected_license_plates.pop(event_id)
if updated_db:
try:
event: Event = Event.get(Event.id == event_id)
except DoesNotExist:
continue
# Skip the event if not an object
if event.data.get("type") != "object":
continue
# Extract valid thumbnail
thumbnail = base64.b64decode(event.thumbnail)
# Embed the thumbnail
self._embed_thumbnail(event_id, thumbnail)
if (
camera_config.genai.enabled
and self.genai_client is not None
and event.data.get("description") is None
and (
not camera_config.genai.objects
or event.label in camera_config.genai.objects
)
and (
not camera_config.genai.required_zones
or set(event.zones) & set(camera_config.genai.required_zones)
)
):
if event.has_snapshot and camera_config.genai.use_snapshot:
with open(
os.path.join(CLIPS_DIR, f"{event.camera}-{event.id}.jpg"),
"rb",
) as image_file:
snapshot_image = image_file.read()
img = cv2.imdecode(
np.frombuffer(snapshot_image, dtype=np.int8),
cv2.IMREAD_COLOR,
)
# crop snapshot based on region before sending off to genai
height, width = img.shape[:2]
x1_rel, y1_rel, width_rel, height_rel = event.data["region"]
x1, y1 = int(x1_rel * width), int(y1_rel * height)
cropped_image = img[
y1 : y1 + int(height_rel * height),
x1 : x1 + int(width_rel * width),
]
_, buffer = cv2.imencode(".jpg", cropped_image)
snapshot_image = buffer.tobytes()
embed_image = (
[snapshot_image]
if event.has_snapshot and camera_config.genai.use_snapshot
else (
[thumbnail for data in self.tracked_events[event_id]]
if len(self.tracked_events.get(event_id, [])) > 0
else [thumbnail]
)
)
# Generate the description. Call happens in a thread since it is network bound.
threading.Thread(
target=self._embed_description,
name=f"_embed_description_{event.id}",
daemon=True,
args=(
event,
embed_image,
),
).start()
# Delete tracked events based on the event_id
if event_id in self.tracked_events:
del self.tracked_events[event_id]
def _process_event_metadata(self):
# Check for regenerate description requests
(topic, event_id, source) = self.event_metadata_subscriber.check_for_update(
timeout=0.01
)
if topic is None:
return
if event_id:
self.handle_regenerate_description(event_id, source)
def _search_face(self, query_embedding: bytes) -> list[tuple[str, float]]:
"""Search for the face most closely matching the embedding."""
sql_query = f"""
SELECT
id,
distance
FROM vec_faces
WHERE face_embedding MATCH ?
AND k = {REQUIRED_FACES} ORDER BY distance
"""
return self.embeddings.db.execute_sql(sql_query, [query_embedding]).fetchall()
def _detect_face(self, input: np.ndarray) -> tuple[int, int, int, int]:
"""Detect faces in input image."""
self.face_detector.setInputSize((input.shape[1], input.shape[0]))
faces = self.face_detector.detect(input)
if faces[1] is None:
return None
face = None
for _, potential_face in enumerate(faces[1]):
raw_bbox = potential_face[0:4].astype(np.uint16)
x: int = max(raw_bbox[0], 0)
y: int = max(raw_bbox[1], 0)
w: int = raw_bbox[2]
h: int = raw_bbox[3]
bbox = (x, y, x + w, y + h)
if face is None or area(bbox) > area(face):
face = bbox
return face
def _process_face(self, obj_data: dict[str, any], frame: np.ndarray) -> None:
"""Look for faces in image."""
id = obj_data["id"]
# don't run for non person objects
if obj_data.get("label") != "person":
logger.debug("Not a processing face for non person object.")
return
# don't overwrite sub label for objects that have a sub label
# that is not a face
if obj_data.get("sub_label") and id not in self.detected_faces:
logger.debug(
f"Not processing face due to existing sub label: {obj_data.get('sub_label')}."
)
return
face: Optional[dict[str, any]] = None
if self.requires_face_detection:
logger.debug("Running manual face detection.")
person_box = obj_data.get("box")
if not person_box:
return None
rgb = cv2.cvtColor(frame, cv2.COLOR_YUV2RGB_I420)
left, top, right, bottom = person_box
person = rgb[top:bottom, left:right]
face = self._detect_face(person)
if not face:
logger.debug("Detected no faces for person object.")
return
face_frame = person[face[1] : face[3], face[0] : face[2]]
face_frame = cv2.cvtColor(face_frame, cv2.COLOR_RGB2BGR)
else:
# don't run for object without attributes
if not obj_data.get("current_attributes"):
logger.debug("No attributes to parse.")
return
attributes: list[dict[str, any]] = obj_data.get("current_attributes", [])
for attr in attributes:
if attr.get("label") != "face":
continue
if face is None or attr.get("score", 0.0) > face.get("score", 0.0):
face = attr
# no faces detected in this frame
if not face:
return
face_box = face.get("box")
# check that face is valid
if not face_box or area(face_box) < self.config.face_recognition.min_area:
logger.debug(f"Invalid face box {face}")
return
face_frame = cv2.cvtColor(frame, cv2.COLOR_YUV2BGR_I420)
face_frame = face_frame[
face_box[1] : face_box[3], face_box[0] : face_box[2]
]
ret, webp = cv2.imencode(
".webp", face_frame, [int(cv2.IMWRITE_WEBP_QUALITY), 100]
)
if not ret:
logger.debug("Not processing face due to error creating cropped image.")
return
embedding = self.embeddings.embed_face("unknown", webp.tobytes(), upsert=False)
query_embedding = serialize(embedding)
best_faces = self._search_face(query_embedding)
logger.debug(f"Detected best faces for person as: {best_faces}")
if not best_faces or len(best_faces) < REQUIRED_FACES:
logger.debug(f"{len(best_faces)} < {REQUIRED_FACES} min required faces.")
return
sub_label = str(best_faces[0][0]).split("-")[0]
avg_score = 0
for face in best_faces:
score = 1.0 - face[1]
if face[0].split("-")[0] != sub_label:
logger.debug("Detected multiple faces, result is not valid.")
return
avg_score += score
avg_score = round(avg_score / REQUIRED_FACES, 2)
if avg_score < self.config.face_recognition.threshold or (
id in self.detected_faces and avg_score <= self.detected_faces[id]
):
logger.debug(
f"Recognized face score {avg_score} is less than threshold ({self.config.face_recognition.threshold}) / previous face score ({self.detected_faces.get(id)})."
)
return
resp = requests.post(
f"{FRIGATE_LOCALHOST}/api/events/{id}/sub_label",
json={
"camera": obj_data.get("camera"),
"subLabel": sub_label,
"subLabelScore": avg_score,
},
)
if resp.status_code == 200:
self.detected_faces[id] = avg_score
def _detect_license_plate(self, input: np.ndarray) -> tuple[int, int, int, int]:
"""Return the dimensions of the input image as [x, y, width, height]."""
height, width = input.shape[:2]
return (0, 0, width, height)
def _process_license_plate(
self, obj_data: dict[str, any], frame: np.ndarray
) -> None:
"""Look for license plates in image."""
id = obj_data["id"]
# don't run for non car objects
if obj_data.get("label") != "car":
logger.debug("Not a processing license plate for non car object.")
return
# don't run for stationary car objects
if obj_data.get("stationary") == True:
logger.debug("Not a processing license plate for a stationary car object.")
return
# don't overwrite sub label for objects that have a sub label
# that is not a license plate
if obj_data.get("sub_label") and id not in self.detected_license_plates:
logger.debug(
f"Not processing license plate due to existing sub label: {obj_data.get('sub_label')}."
)
return
license_plate: Optional[dict[str, any]] = None
if self.requires_license_plate_detection:
logger.debug("Running manual license_plate detection.")
car_box = obj_data.get("box")
if not car_box:
return None
rgb = cv2.cvtColor(frame, cv2.COLOR_YUV2RGB_I420)
left, top, right, bottom = car_box
car = rgb[top:bottom, left:right]
license_plate = self._detect_license_plate(car)
if not license_plate:
logger.debug("Detected no license plates for car object.")
return
license_plate_frame = car[
license_plate[1] : license_plate[3], license_plate[0] : license_plate[2]
]
license_plate_frame = cv2.cvtColor(license_plate_frame, cv2.COLOR_RGB2BGR)
else:
# don't run for object without attributes
if not obj_data.get("current_attributes"):
logger.debug("No attributes to parse.")
return
attributes: list[dict[str, any]] = obj_data.get("current_attributes", [])
for attr in attributes:
if attr.get("label") != "license_plate":
continue
if license_plate is None or attr.get("score", 0.0) > license_plate.get(
"score", 0.0
):
license_plate = attr
# no license plates detected in this frame
if not license_plate:
return
license_plate_box = license_plate.get("box")
# check that license plate is valid
if (
not license_plate_box
or area(license_plate_box) < self.config.lpr.min_area
):
logger.debug(f"Invalid license plate box {license_plate}")
return
license_plate_frame = cv2.cvtColor(frame, cv2.COLOR_YUV2BGR_I420)
license_plate_frame = license_plate_frame[
license_plate_box[1] : license_plate_box[3],
license_plate_box[0] : license_plate_box[2],
]
# run detection, returns results sorted by confidence, best first
license_plates, confidences, areas = (
self.license_plate_recognition.process_license_plate(license_plate_frame)
)
logger.debug(f"Text boxes: {license_plates}")
logger.debug(f"Confidences: {confidences}")
logger.debug(f"Areas: {areas}")
if license_plates:
for plate, confidence, text_area in zip(license_plates, confidences, areas):
avg_confidence = (
(sum(confidence) / len(confidence)) if confidence else 0
)
logger.debug(
f"Detected text: {plate} (average confidence: {avg_confidence:.2f}, area: {text_area} pixels)"
)
else:
# no plates found
logger.debug("No text detected")
return
top_plate, top_char_confidences = license_plates[0], confidences[0]
avg_confidence = (
(sum(top_char_confidences) / len(top_char_confidences))
if top_char_confidences
else 0
)
# Check if we have a previously detected plate for this ID
if id in self.detected_license_plates:
prev_plate = self.detected_license_plates[id]["plate"]
prev_char_confidences = self.detected_license_plates[id]["char_confidences"]
prev_avg_confidence = (
(sum(prev_char_confidences) / len(prev_char_confidences))
if prev_char_confidences
else 0
)
# Define conditions for keeping the previous plate
shorter_than_previous = len(top_plate) < len(prev_plate)
lower_avg_confidence = avg_confidence <= prev_avg_confidence
# Compare character-by-character confidence where possible
min_length = min(len(top_plate), len(prev_plate))
char_confidence_comparison = sum(
1
for i in range(min_length)
if top_char_confidences[i] <= prev_char_confidences[i]
)
worse_char_confidences = char_confidence_comparison >= min_length / 2
if shorter_than_previous or (
lower_avg_confidence and worse_char_confidences
):
logger.debug(
f"Keeping previous plate. New plate stats: "
f"length={len(top_plate)}, avg_conf={avg_confidence:.2f} "
f"vs Previous: length={len(prev_plate)}, avg_conf={prev_avg_confidence:.2f}"
)
return
# Check against minimum confidence threshold
if avg_confidence < self.lpr_config.threshold:
logger.debug(
f"Average confidence {avg_confidence} is less than threshold ({self.lpr_config.threshold})"
)
return
# Determine subLabel based on known plates
# Default to the detected plate, use label name if there's a match
sub_label = top_plate
for label, plates in self.lpr_config.known_plates.items():
if top_plate in plates:
sub_label = label
break
# Send the result to the API
resp = requests.post(
f"{FRIGATE_LOCALHOST}/api/events/{id}/sub_label",
json={
"camera": obj_data.get("camera"),
"subLabel": sub_label,
"subLabelScore": avg_confidence,
},
)
if resp.status_code == 200:
self.detected_license_plates[id] = {
"plate": top_plate,
"char_confidences": top_char_confidences,
}
def _create_thumbnail(self, yuv_frame, box, height=500) -> Optional[bytes]:
"""Return jpg thumbnail of a region of the frame."""
frame = cv2.cvtColor(yuv_frame, cv2.COLOR_YUV2BGR_I420)
region = calculate_region(
frame.shape, box[0], box[1], box[2], box[3], height, multiplier=1.4
)
frame = frame[region[1] : region[3], region[0] : region[2]]
width = int(height * frame.shape[1] / frame.shape[0])
frame = cv2.resize(frame, dsize=(width, height), interpolation=cv2.INTER_AREA)
ret, jpg = cv2.imencode(".jpg", frame, [int(cv2.IMWRITE_JPEG_QUALITY), 100])
if ret:
return jpg.tobytes()
return None
def _embed_thumbnail(self, event_id: str, thumbnail: bytes) -> None:
"""Embed the thumbnail for an event."""
self.embeddings.embed_thumbnail(event_id, thumbnail)
def _embed_description(self, event: Event, thumbnails: list[bytes]) -> None:
"""Embed the description for an event."""
camera_config = self.config.cameras[event.camera]
description = self.genai_client.generate_description(
camera_config, thumbnails, event
)
if not description:
logger.debug("Failed to generate description for %s", event.id)
return
# fire and forget description update
self.requestor.send_data(
UPDATE_EVENT_DESCRIPTION,
{"id": event.id, "description": description},
)
# Embed the description
self.embeddings.embed_description(event.id, description)
logger.debug(
"Generated description for %s (%d images): %s",
event.id,
len(thumbnails),
description,
)
def handle_regenerate_description(self, event_id: str, source: str) -> None:
try:
event: Event = Event.get(Event.id == event_id)
except DoesNotExist:
logger.error(f"Event {event_id} not found for description regeneration")
return
camera_config = self.config.cameras[event.camera]
if not camera_config.genai.enabled or self.genai_client is None:
logger.error(f"GenAI not enabled for camera {event.camera}")
return
thumbnail = base64.b64decode(event.thumbnail)
logger.debug(
f"Trying {source} regeneration for {event}, has_snapshot: {event.has_snapshot}"
)
if event.has_snapshot and source == "snapshot":
with open(
os.path.join(CLIPS_DIR, f"{event.camera}-{event.id}.jpg"),
"rb",
) as image_file:
snapshot_image = image_file.read()
img = cv2.imdecode(
np.frombuffer(snapshot_image, dtype=np.int8), cv2.IMREAD_COLOR
)
# crop snapshot based on region before sending off to genai
height, width = img.shape[:2]
x1_rel, y1_rel, width_rel, height_rel = event.data["region"]
x1, y1 = int(x1_rel * width), int(y1_rel * height)
cropped_image = img[
y1 : y1 + int(height_rel * height), x1 : x1 + int(width_rel * width)
]
_, buffer = cv2.imencode(".jpg", cropped_image)
snapshot_image = buffer.tobytes()
embed_image = (
[snapshot_image]
if event.has_snapshot and source == "snapshot"
else (
[thumbnail for data in self.tracked_events[event_id]]
if len(self.tracked_events.get(event_id, [])) > 0
else [thumbnail]
)
)
self._embed_description(event, embed_image)