fix: reset counter after each loop

ref: N25B-395

src/control_backend/agents/bdi/agentspeak_generator.py

@@ -30,6 +30,7 @@ from control_backend.schemas.program import (
     BasicNorm,
     ConditionalNorm,
     EmotionBelief,
+    FaceBelief,
     GestureAction,
     Goal,
     InferredBelief,
@@ -682,11 +683,15 @@ class AgentSpeakGenerator:
         :return: An AstLiteral representing the semantic belief.
         """
         return AstLiteral(self.slugify(sb))
-    
+
     @_astify.register
     def _(self, eb: EmotionBelief) -> AstExpression:
         return AstLiteral("emotion_detected", [AstAtom(eb.emotion)])
 
+    @_astify.register
+    def _(self, fb: FaceBelief) -> AstExpression:
+        return AstLiteral("face_present")
+
     @_astify.register
     def _(self, ib: InferredBelief) -> AstExpression:
         """

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

@@ -14,7 +14,7 @@ from control_backend.agents.perception.visual_emotion_recognition_agent.visual_e
 )
 from control_backend.core.agent_system import InternalMessage
 from control_backend.core.config import settings
-from control_backend.schemas.belief_message import Belief
+from control_backend.schemas.belief_message import Belief, BeliefMessage
 
 
 class VisualEmotionRecognitionAgent(BaseAgent):
@@ -44,6 +44,7 @@ class VisualEmotionRecognitionAgent(BaseAgent):
         self.timeout_ms = timeout_ms
         self.window_duration = window_duration
         self.min_frames_required = min_frames_required
+        self._face_detected = False
 
         # Pause functionality
         # NOTE: flag is set when running, cleared when paused
@@ -89,6 +90,9 @@ class VisualEmotionRecognitionAgent(BaseAgent):
         # Tracks counts of detected emotions per face index
         face_stats = defaultdict(Counter)
 
+        # How many times a face has been detected
+        face_detection_yes_no = [0, 0]
+
         prev_dominant_emotions = set()
 
         while self._running:
@@ -97,8 +101,8 @@ class VisualEmotionRecognitionAgent(BaseAgent):
 
                 width, height, image_bytes = await self.video_in_socket.recv_multipart()
 
-                width = int.from_bytes(width, 'little')
-                height = int.from_bytes(height, 'little')
+                width = int.from_bytes(width, "little")
+                height = int.from_bytes(height, "little")
 
                 # Convert bytes to a numpy buffer
                 image_array = np.frombuffer(image_bytes, np.uint8)
@@ -107,6 +111,13 @@ class VisualEmotionRecognitionAgent(BaseAgent):
 
                 # Get the dominant emotion from each face
                 current_emotions = self.emotion_recognizer.sorted_dominant_emotions(frame)
+
+                # Update face face_detection_yes_no
+                if len(current_emotions) > 0:
+                    face_detection_yes_no[0] += 1
+                else:
+                    face_detection_yes_no[1] += 1
+
                 # Update emotion counts for each detected face
                 for i, emotion in enumerate(current_emotions):
                     face_stats[i][emotion] += 1
@@ -122,6 +133,20 @@ class VisualEmotionRecognitionAgent(BaseAgent):
                             dominant_emotion = counter.most_common(1)[0][0]
                             window_dominant_emotions.add(dominant_emotion)
 
+                    if (
+                        face_detection_yes_no[0] > face_detection_yes_no[1]
+                        and not self._face_detected
+                    ):
+                        self._face_detected = True
+                        await self._inform_face_detected()
+                    elif (
+                        face_detection_yes_no[0] <= face_detection_yes_no[1] and self._face_detected
+                    ):
+                        self._face_detected = False
+                        await self._inform_face_detected()
+
+                    face_detection_yes_no = [0, 0]
+
                     await self.update_emotions(prev_dominant_emotions, window_dominant_emotions)
                     prev_dominant_emotions = window_dominant_emotions
                     face_stats.clear()
@@ -133,7 +158,6 @@ class VisualEmotionRecognitionAgent(BaseAgent):
             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]):
         """
         Compare emotions from previous window and current emotions,
@@ -149,9 +173,7 @@ class VisualEmotionRecognitionAgent(BaseAgent):
         for emotion in emotions_to_remove:
             self.logger.info(f"Emotion '{emotion}' has disappeared.")
             try:
-                emotion_beliefs_remove.append(
-                    Belief(name="emotion_detected", arguments=[emotion], remove=True)
-                )
+                emotion_beliefs_remove.append(Belief(name="emotion_detected", arguments=[emotion]))
             except ValidationError:
                 self.logger.warning("Invalid belief for emotion removal: %s", emotion)
 
@@ -175,11 +197,25 @@ class VisualEmotionRecognitionAgent(BaseAgent):
         )
         await self.send(message)
 
+    async def _inform_face_detected(self):
+        if self._face_detected:
+            belief_message = BeliefMessage(create=[Belief(name="face_present")])
+        else:
+            belief_message = BeliefMessage(delete=[Belief(name="face_present")])
+
+        msg = InternalMessage(
+            to=settings.agent_settings.bdi_core_name,
+            thread="beliefs",
+            body=belief_message.model_dump_json(),
+        )
+
+        await self.send(msg)
+
     async def handle_message(self, msg: InternalMessage):
         """
         Handle incoming messages.
 
-        Expects messages to pause or resume the Visual Emotion Recognition 
+        Expects messages to pause or resume the Visual Emotion Recognition
         processing from User Interrupt Agent.
 
         :param msg: The received internal message.
@@ -204,4 +240,3 @@ class VisualEmotionRecognitionAgent(BaseAgent):
         """
         self.video_in_socket.close()
         await super().stop()
-        

src/control_backend/core/config.py

@@ -82,7 +82,7 @@ class BehaviourSettings(BaseModel):
     :ivar transcription_words_per_token: Estimated words per token for transcription timing.
     :ivar transcription_token_buffer: Buffer for transcription tokens.
     :ivar conversation_history_length_limit: The maximum amount of messages to extract beliefs from.
-    :ivar visual_emotion_recognition_window_duration_s: Duration in seconds over which to aggregate 
+    :ivar visual_emotion_recognition_window_duration_s: Duration in seconds over which to aggregate
     emotions and update emotion beliefs.
     :ivar visual_emotion_recognition_min_frames_per_face: Minimum number of frames per face required
     to consider a face valid.
@@ -112,7 +112,7 @@ class BehaviourSettings(BaseModel):
     conversation_history_length_limit: int = 10
 
     # Visual Emotion Recognition settings
-    visual_emotion_recognition_window_duration_s: int = 5
+    visual_emotion_recognition_window_duration_s: int = 3
     visual_emotion_recognition_min_frames_per_face: int = 3
     # AgentSpeak related settings
     trigger_time_to_wait: int = 2000

src/control_backend/schemas/program.py

@@ -41,8 +41,8 @@ class LogicalOperator(Enum):
     OR = "OR"
 
 
-type Belief = KeywordBelief | SemanticBelief | InferredBelief | EmotionBelief
-type BasicBelief = KeywordBelief | SemanticBelief | EmotionBelief
+type Belief = KeywordBelief | SemanticBelief | InferredBelief | EmotionBelief | FaceBelief
+type BasicBelief = KeywordBelief | SemanticBelief | EmotionBelief | FaceBelief
 
 
 class KeywordBelief(ProgramElement):
@@ -105,6 +105,7 @@ class InferredBelief(ProgramElement):
     left: Belief
     right: Belief
 
+
 class EmotionBelief(ProgramElement):
     """
     Represents a belief that is set when a certain emotion is detected.
@@ -115,6 +116,16 @@ class EmotionBelief(ProgramElement):
     name: str = ""
     emotion: str
 
+
+class FaceBelief(ProgramElement):
+    """
+    Represents the belief that at least one face is currently in view.
+    """
+
+    name: str = ""
+    face_present: bool
+
+
 class Norm(ProgramElement):
     """
     Base class for behavioral norms that guide the robot's interactions.
@@ -329,4 +340,4 @@ class Program(BaseModel):
 if __name__ == "__main__":
     input = input("Enter program JSON: ")
     program = Program.model_validate_json(input)
-    print(program)
+    print(program)