src/control_backend/agents/perception/visual_emotion_recognition_agent/visual_emotion_recognition_agent.py

@@ -64,6 +64,8 @@ class VisualEmotionRecognitionAgent(BaseAgent):
 
         self.video_in_socket = azmq.Context.instance().socket(zmq.SUB)
 
+        self.video_in_socket.setsockopt(zmq.RCVHWM, 3)
+
         if self.socket_bind:
             self.video_in_socket.bind(self.socket_address)
         else:
@@ -71,7 +73,6 @@ class VisualEmotionRecognitionAgent(BaseAgent):
 
         self.video_in_socket.setsockopt_string(zmq.SUBSCRIBE, "")
         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())
 
@@ -95,21 +96,18 @@ class VisualEmotionRecognitionAgent(BaseAgent):
             try:
                 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
-                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 = 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
+                frame = image_array.reshape((height, width, 3))
 
                 # 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
                 for i, emotion in enumerate(current_emotions):
                     face_stats[i][emotion] += 1
@@ -133,6 +131,9 @@ class VisualEmotionRecognitionAgent(BaseAgent):
             except zmq.Again:
                 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]):
         """