pepperplus/pepperplus-cb
1
0
Fork 0
Code Issues Pull Requests 1 Actions Packages Projects Releases Wiki Activity

refactor: visual emotion recognition agent now receives image in three parts: width, height and image bytes

Browse Source 
ref: N25B-393
This commit is contained in:
Storm
2026-01-29 12:00:14 +01:00
parent 0b1c2ce20a
commit 35ffd33968
1 changed files with 13 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
src/control_backend/agents/perception/visual_emotion_recognition_agent/visual_emotion_recognition_agent.py
View File

@@ -64,6 +64,8 @@ class VisualEmotionRecognitionAgent(BaseAgent):
self.video_in_socket = azmq.Context.instance().socket(zmq.SUB) self.video_in_socket = azmq.Context.instance().socket(zmq.SUB)
self.video_in_socket.setsockopt(zmq.RCVHWM, 3)
if self.socket_bind: if self.socket_bind:
self.video_in_socket.bind(self.socket_address) self.video_in_socket.bind(self.socket_address)
else: else:
@@ -71,7 +73,6 @@ class VisualEmotionRecognitionAgent(BaseAgent):
self.video_in_socket.setsockopt_string(zmq.SUBSCRIBE, "") self.video_in_socket.setsockopt_string(zmq.SUBSCRIBE, "")
self.video_in_socket.setsockopt(zmq.RCVTIMEO, self.timeout_ms) self.video_in_socket.setsockopt(zmq.RCVTIMEO, self.timeout_ms)
self.video_in_socket.setsockopt(zmq.CONFLATE, 1)
self.add_behavior(self.emotion_update_loop()) self.add_behavior(self.emotion_update_loop())
@@ -95,21 +96,18 @@ class VisualEmotionRecognitionAgent(BaseAgent):
try: try:
await self._paused.wait() await self._paused.wait()
frame_bytes = await self.video_in_socket.recv() width, height, image_bytes = await self.video_in_socket.recv_multipart()
width = int.from_bytes(width, 'little')
height = int.from_bytes(height, 'little')
# Convert bytes to a numpy buffer # Convert bytes to a numpy buffer
nparr = np.frombuffer(frame_bytes, np.uint8) image_array = np.frombuffer(image_bytes, np.uint8)
# Decode image into the generic Numpy Array DeepFace expects frame = image_array.reshape((height, width, 3))
frame_image = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
if frame_image is None:
# Could not decode image, skip this frame
self.logger.warning("Received invalid video frame, skipping.")
continue
# Get the dominant emotion from each face # Get the dominant emotion from each face
current_emotions = self.emotion_recognizer.sorted_dominant_emotions(frame_image) current_emotions = self.emotion_recognizer.sorted_dominant_emotions(frame)
# Update emotion counts for each detected face # Update emotion counts for each detected face
for i, emotion in enumerate(current_emotions): for i, emotion in enumerate(current_emotions):
face_stats[i][emotion] += 1 face_stats[i][emotion] += 1
@@ -133,6 +131,9 @@ class VisualEmotionRecognitionAgent(BaseAgent):
except zmq.Again: except zmq.Again:
self.logger.warning("No video frame received within timeout.") self.logger.warning("No video frame received within timeout.")
except Exception as e:
self.logger.error(f"Error in emotion recognition loop: {e}")
async def update_emotions(self, prev_emotions: set[str], emotions: set[str]): async def update_emotions(self, prev_emotions: set[str], emotions: set[str]):
""" """