feat: fully working face detection
ref: N25B-397
This commit is contained in:
@@ -17,6 +17,8 @@ class AgentSettings(object):
|
||||
:vartype video_sender_port: int
|
||||
:ivar audio_sender_port: Port used for sending audio data, defaults to 5558.
|
||||
:vartype audio_sender_port: int
|
||||
:ivar face_detection_port: Port used for sending face detection events, defaults to 5559.
|
||||
:vartype face_detection_port: int
|
||||
"""
|
||||
def __init__(
|
||||
self,
|
||||
@@ -25,12 +27,14 @@ class AgentSettings(object):
|
||||
main_receiver_port=None,
|
||||
video_sender_port=None,
|
||||
audio_sender_port=None,
|
||||
face_detection_port=None,
|
||||
):
|
||||
self.control_backend_host = get_config(control_backend_host, "AGENT__CONTROL_BACKEND_HOST", "localhost")
|
||||
self.actuation_receiver_port = get_config(actuation_receiver_port, "AGENT__ACTUATION_RECEIVER_PORT", 5557, int)
|
||||
self.main_receiver_port = get_config(main_receiver_port, "AGENT__MAIN_RECEIVER_PORT", 5555, int)
|
||||
self.video_sender_port = get_config(video_sender_port, "AGENT__VIDEO_SENDER_PORT", 5556, int)
|
||||
self.audio_sender_port = get_config(audio_sender_port, "AGENT__AUDIO_SENDER_PORT", 5558, int)
|
||||
self.face_detection_port = get_config(face_detection_port, "AGENT__FACE_DETECTION_PORT", 5559, int)
|
||||
|
||||
|
||||
class VideoConfig(object):
|
||||
|
||||
@@ -1,52 +1,49 @@
|
||||
from __future__ import unicode_literals
|
||||
|
||||
import logging
|
||||
import threading
|
||||
import zmq
|
||||
import time
|
||||
|
||||
from robot_interface.endpoints.socket_base import SocketBase
|
||||
from robot_interface.state import state
|
||||
from robot_interface.core.config import settings
|
||||
|
||||
|
||||
class FaceDetectionSender(SocketBase):
|
||||
"""
|
||||
Face detection sender endpoint.
|
||||
Minimal face detection sender with a shared flag.
|
||||
|
||||
Subscribes to ALFaceDetection and forwards face detection
|
||||
events over ZeroMQ.
|
||||
Polls ALMemory["FaceDetected"] and keeps a simple boolean
|
||||
indicating if a face is currently detected.
|
||||
"""
|
||||
|
||||
def __init__(self, zmq_context, port=settings.agent_settings.face_sender_port):
|
||||
def __init__(self, zmq_context, port=None):
|
||||
super(FaceDetectionSender, self).__init__("face")
|
||||
self.create_socket(zmq_context, zmq.PUB, port)
|
||||
|
||||
if port:
|
||||
self.create_socket(zmq_context, None, port) # PUB not used here
|
||||
self._face_service = None
|
||||
self._memory_service = None
|
||||
self._subscriber = None
|
||||
self._thread = None
|
||||
|
||||
# Shared status for MainReceiver polling
|
||||
self.face_detected = False
|
||||
self._last_seen_face_time = 0
|
||||
|
||||
def start_face_detection(self):
|
||||
"""
|
||||
Initializes ALFaceDetection and starts listening for face events.
|
||||
"""
|
||||
if not state.qi_session:
|
||||
logging.info("No Qi session available. Not starting face detection.")
|
||||
return
|
||||
|
||||
import qi # Lazy import (same pattern as rest of codebase)
|
||||
|
||||
import qi
|
||||
self._face_service = state.qi_session.service("ALFaceDetection")
|
||||
self._memory_service = state.qi_session.service("ALMemory")
|
||||
|
||||
# Enable face detection
|
||||
self._face_service.setTrackingEnabled(True)
|
||||
# Enable minimal detection
|
||||
self._face_service.setTrackingEnabled(False)
|
||||
self._face_service.setRecognitionEnabled(False)
|
||||
|
||||
# Subscribe to FaceDetected memory event
|
||||
self._subscriber = self._memory_service.subscriber("FaceDetected")
|
||||
self._subscriber.signal.connect(self._on_face_detected)
|
||||
# Required to activate extractor
|
||||
self._face_service.subscribe("FaceDetectionSender", 500, 0.0)
|
||||
|
||||
# Start keep-alive thread
|
||||
self._thread = threading.Thread(target=self._face_loop)
|
||||
self._thread.daemon = True
|
||||
self._thread.start()
|
||||
@@ -58,56 +55,33 @@ class FaceDetectionSender(SocketBase):
|
||||
Keeps the face detection alive until shutdown.
|
||||
"""
|
||||
while not state.exit_event.is_set():
|
||||
state.exit_event.wait(0.1)
|
||||
try:
|
||||
value = self._memory_service.getData("FaceDetected", 0)
|
||||
face_present = (
|
||||
value
|
||||
and len(value) > 1
|
||||
and value[1]
|
||||
and value[1][0]
|
||||
and len(value[1][0]) > 0
|
||||
)
|
||||
|
||||
self._cleanup()
|
||||
now = time.time()
|
||||
if face_present:
|
||||
self._last_seen_face_time = now
|
||||
|
||||
def _on_face_detected(self, value):
|
||||
"""
|
||||
Callback for ALMemory FaceDetected event.
|
||||
# Consider face "lost" after 3s
|
||||
self.face_detected = (now - self._last_seen_face_time) < 3
|
||||
|
||||
:param value: Face detection data structure from NAOqi
|
||||
"""
|
||||
if not value or len(value) < 2:
|
||||
return
|
||||
except Exception:
|
||||
logging.exception("Error reading FaceDetected")
|
||||
|
||||
timestamp = value[0]
|
||||
faces = value[1]
|
||||
|
||||
face_data = {
|
||||
"timestamp": timestamp,
|
||||
"face_count": len(faces),
|
||||
"faces": []
|
||||
}
|
||||
|
||||
for face in faces:
|
||||
face_info = face[0] # Shape info
|
||||
extra_info = face[1] # Extra info (ID, score, etc.)
|
||||
|
||||
face_data["faces"].append({
|
||||
"alpha": face_info[1],
|
||||
"beta": face_info[2],
|
||||
"width": face_info[3],
|
||||
"height": face_info[4],
|
||||
"confidence": extra_info[1] if len(extra_info) > 1 else None
|
||||
})
|
||||
time.sleep(0.1)
|
||||
|
||||
def stop_face_detection(self):
|
||||
try:
|
||||
self.socket.send_json(face_data)
|
||||
except Exception:
|
||||
logging.warn("Failed to send face detection data.")
|
||||
|
||||
def _cleanup(self):
|
||||
"""
|
||||
Cleanup subscriptions and disable face detection.
|
||||
"""
|
||||
try:
|
||||
if self._subscriber:
|
||||
self._subscriber.signal.disconnect(self._on_face_detected)
|
||||
|
||||
if self._face_service:
|
||||
self._face_service.unsubscribe("FaceDetectionSender")
|
||||
self._face_service.setTrackingEnabled(False)
|
||||
|
||||
logging.info("Face detection stopped.")
|
||||
except Exception:
|
||||
logging.warn("Error during face detection cleanup.")
|
||||
logging.warning("Error during face detection cleanup.")
|
||||
|
||||
@@ -4,6 +4,7 @@ from robot_interface.endpoints.receiver_base import ReceiverBase
|
||||
from robot_interface.state import state
|
||||
|
||||
from robot_interface.core.config import settings
|
||||
from robot_interface.endpoints.face_detector import FaceDetectionSender
|
||||
|
||||
|
||||
class MainReceiver(ReceiverBase):
|
||||
@@ -36,6 +37,29 @@ class MainReceiver(ReceiverBase):
|
||||
:rtype: dict[str, str | list[dict]]
|
||||
"""
|
||||
return {"endpoint": "ping", "data": message.get("data")}
|
||||
|
||||
@staticmethod
|
||||
def _handle_face(message):
|
||||
"""
|
||||
Handles sending face data to the cb
|
||||
Sends if it sees a face or not
|
||||
|
||||
:param message: face data.
|
||||
:type message: int
|
||||
|
||||
:return: A response to CB containing the amount of faces
|
||||
:rtype: int
|
||||
"""
|
||||
# Poll the FaceDetectionSender status
|
||||
face_sender = next(
|
||||
(s for s in state.sockets if isinstance(s, FaceDetectionSender)),
|
||||
None
|
||||
)
|
||||
if face_sender:
|
||||
return {"endpoint": "face", "data": face_sender.face_detected}
|
||||
else:
|
||||
return {"endpoint": "face", "data": False}
|
||||
|
||||
|
||||
@staticmethod
|
||||
def _handle_port_negotiation(message):
|
||||
@@ -86,6 +110,8 @@ class MainReceiver(ReceiverBase):
|
||||
"""
|
||||
if message["endpoint"] == "ping":
|
||||
return self._handle_ping(message)
|
||||
elif message["endpoint"] == "face":
|
||||
return self._handle_face(message)
|
||||
elif message["endpoint"].startswith("negotiate"):
|
||||
return self._handle_negotiation(message)
|
||||
|
||||
|
||||
@@ -12,6 +12,8 @@ from robot_interface.endpoints.video_sender import VideoSender
|
||||
from robot_interface.state import state
|
||||
from robot_interface.core.config import settings
|
||||
from robot_interface.utils.timeblock import TimeBlock
|
||||
from robot_interface.endpoints.face_detector import FaceDetectionSender
|
||||
|
||||
|
||||
|
||||
def main_loop(context):
|
||||
@@ -35,6 +37,12 @@ def main_loop(context):
|
||||
video_sender.start_video_rcv()
|
||||
audio_sender.start()
|
||||
|
||||
# --- Face detection sender ---
|
||||
face_sender = FaceDetectionSender(context)
|
||||
state.sockets.append(face_sender)
|
||||
face_sender.start_face_detection()
|
||||
|
||||
|
||||
# Sockets that can run on the main thread. These sockets' endpoints should not block for long (say 50 ms at most).
|
||||
receivers = [main_receiver, actuation_receiver]
|
||||
|
||||
|
||||
Reference in New Issue
Block a user