Update pages/Camera.py

pages/Camera.py

@@ -6,11 +6,25 @@ import asyncio
 import streamlit as st
 from streamlit_webrtc import WebRtcMode, webrtc_streamer
 import av
-
 from utils import SLInference
 
 logger = logging.getLogger(__name__)
 
+# Callback function to process video frames
+def video_frame_callback(frame: av.VideoFrame, inference_thread, gestures_deque):
+    img = frame.to_ndarray(format="rgb24")
+
+    inference_thread.input_queue.append(img)
+    gesture = inference_thread.pred
+
+    if gesture not in ['no', '']:
+        if not gestures_deque:
+            gestures_deque.append(gesture)
+        elif gesture != gestures_deque[-1]:
+            gestures_deque.append(gesture)
+
+    return av.VideoFrame.from_ndarray(img, format="rgb24")
+
 def main(config_path):
     """
     Main function of the app.
@@ -21,47 +35,42 @@ def main(config_path):
     gestures_deque = deque(maxlen=5)
 
     # Set up Streamlit interface
-    st.set_page_config(page_title="Gesture Recognition", layout="wide")
     st.title("Sign Language Recognition Demo")
+    image_place = st.empty()
     text_output = st.empty()
     last_5_gestures = st.empty()
     st.markdown(
         """
         This application is designed to recognize sign language using a webcam feed.
         The model has been trained to recognize various sign language gestures and display the corresponding text in real-time.
+        
         The project is open for collaboration. If you have any suggestions or want to contribute, please feel free to reach out.
         """
     )
 
-    def video_frame_callback(frame: av.VideoFrame) -> av.VideoFrame:
-        img_rgb = frame.to_ndarray(format="rgb24")
-        inference_thread.input_queue.append(cv2.resize(img_rgb, (224, 224)))
-        return frame
-
     webrtc_ctx = webrtc_streamer(
         key="sign-language-recognition",
         mode=WebRtcMode.SENDRECV,
-        video_frame_callback=video_frame_callback,
+        video_frame_callback=lambda frame: video_frame_callback(frame, inference_thread, gestures_deque),
         media_stream_constraints={"video": True, "audio": False},
         async_processing=True,
     )
 
     while True:
-        if not webrtc_ctx.state.playing:
-            continue
+        if webrtc_ctx.video_receiver:
+            gesture = inference_thread.pred
 
-        gesture = inference_thread.pred
-        if gesture not in ['no', '']:
-            if not gestures_deque:
-                gestures_deque.append(gesture)
-            elif gesture != gestures_deque[-1]:
-                gestures_deque.append(gesture)
+            if gesture not in ['no', '']:
+                if not gestures_deque:
+                    gestures_deque.append(gesture)
+                elif gesture != gestures_deque[-1]:
+                    gestures_deque.append(gesture)
 
-        text_output.markdown(f'<p style="font-size:20px"> Current gesture: {gesture}</p>',
-                             unsafe_allow_html=True)
-        last_5_gestures.markdown(f'<p style="font-size:20px"> Last 5 gestures: {" ".join(gestures_deque)}</p>',
+            text_output.markdown(f'<p style="font-size:20px"> Current gesture: {gesture}</p>',
                                  unsafe_allow_html=True)
-        print(gestures_deque)
+            last_5_gestures.markdown(f'<p style="font-size:20px"> Last 5 gestures: {" ".join(gestures_deque)}</p>',
+                                     unsafe_allow_html=True)
+            print(gestures_deque)
 
 if __name__ == "__main__":
     asyncio.set_event_loop(asyncio.new_event_loop())