Update pages/Camera.py

pages/Camera.py

@@ -2,13 +2,13 @@ import logging
 import queue
 from collections import deque
 import asyncio
-
+import av
+import cv2
+import numpy as np
 import streamlit as st
 from streamlit_webrtc import WebRtcMode, webrtc_streamer
-
 from utils import SLInference
 
-
 logger = logging.getLogger(__name__)
 
 def main(config_path):
@@ -18,15 +18,10 @@ def main(config_path):
     inference_thread = SLInference(config_path)
     inference_thread.start()
 
-    webrtc_ctx = webrtc_streamer(
-        key="video-sendonly",
-        mode=WebRtcMode.SENDONLY,
-        media_stream_constraints={"video": True},
-    )
-
     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()
@@ -35,22 +30,34 @@ def main(config_path):
         """
         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.
         """
     )
 
+    result_queue = queue.Queue()
+
+    def video_frame_callback(frame: av.VideoFrame) -> av.VideoFrame:
+        img_rgb = frame.to_ndarray(format="rgb24")
+        result_queue.put(img_rgb)
+        return frame
+
+    webrtc_ctx = webrtc_streamer(
+        key="sign-language-recognition",
+        mode=WebRtcMode.SENDRECV,
+        video_frame_callback=video_frame_callback,
+        media_stream_constraints={"video": True, "audio": False},
+        async_processing=True,
+    )
+
     while True:
-        if webrtc_ctx.video_receiver:
-            try:
-                video_frame = webrtc_ctx.video_receiver.get_frame(timeout=1)
-            except queue.Empty:
-                logger.warning("Queue is empty")
-                continue
+        if not webrtc_ctx.state.playing:
+            continue
+
+        if not result_queue.empty():
+            img_rgb = result_queue.get()
 
-            img_rgb = video_frame.to_ndarray(format="rgb24")
             image_place.image(img_rgb)
-            inference_thread.input_queue.append(video_frame.reformat(224,224).to_ndarray(format="rgb24"))
+            inference_thread.input_queue.append(cv2.resize(img_rgb, (224, 224)))
 
             gesture = inference_thread.pred
             if gesture not in ['no', '']:
@@ -62,11 +69,9 @@ def main(config_path):
             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>',
-                                 unsafe_allow_html=True)
+                                     unsafe_allow_html=True)
             print(gestures_deque)
 
-
-
 if __name__ == "__main__":
     asyncio.set_event_loop(asyncio.new_event_loop())
     main("configs/config.json")