changes

app.py

@@ -21,49 +21,33 @@ def main():
     st.title('SpatialSense')
     st.write('Github: https://github.com/kabir12345/SpatialSense')
 
-    # Display CPU usage
-    cpu_usage = psutil.cpu_percent(interval=1)
-    st.metric(label="CPU Usage", value=f"{cpu_usage} %")
-
     # Initialize the depth-estimation pipeline
     pipe = pipeline(task="depth-estimation", model="LiheYoung/depth-anything-small-hf")
 
-    # OpenCV camera initialization
-    cap = cv2.VideoCapture(0)  # 0 is usually the default camera
-    frame_rate = 10  # Frame rate can be adjusted for performance
-    wait_time = 1 / frame_rate
-
-    placeholder = st.empty()
-    text_placeholder = st.empty()
+    # Streamlit-WebRTC component
+    webrtc_ctx = webrtc_streamer(key="example", mode=WebRtcMode.SENDRECV)
 
-    try:
+    if webrtc_ctx.video_receiver:
         while True:
-            ret, frame = cap.read()
-            if not ret:
-                break
+            frame = webrtc_ctx.video_receiver.get_frame(timeout=None)
+            if frame is None:
+                continue
 
-            # Convert the captured frame to PIL Image
-            pil_img = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
+            image = frame.to_ndarray(format="bgr24")
+            pil_img = Image.fromarray(image)
 
-            # Process depth estimation
+            # Perform depth estimation
             depth_mask = apply_depth_estimation(pipe, pil_img)
-            placeholder.image(depth_mask, channels="RGB")
-
-            # Refresh rate control
-            time.sleep(wait_time)
 
-    except KeyboardInterrupt:
-        print("Stopping camera...")
+            # Convert PIL Image to NumPy array for display in Streamlit
+            depth_mask_np = np.array(depth_mask)
 
-    finally:
-        cap.release()
+            # Display the processed image
+            st.image(depth_mask_np, caption="Processed Depth Image", channels="BGR")
 
-def embed_twitch_stream(channel_name):
-    embed_url = f"https://player.twitch.tv/?channel={channel_name}&parent=localhost"
-    html_code = f'<iframe src="{embed_url}" height="394" width="700" frameborder="0" allowfullscreen="true" scrolling="no" allow="autoplay; fullscreen"></iframe>'
-    html(html_code, height=400)
 
 def apply_depth_estimation(pipe, pil_img):
+    # Assume the rest of your depth estimation logic is defined here
     original_width, original_height = pil_img.size
     depth = pipe(pil_img)["depth"]
     depth_tensor = torch.from_numpy(np.array(depth)).unsqueeze(0).unsqueeze(0).float()

requirements.txt

@@ -198,4 +198,5 @@ trio==0.25.0
 trio-websocket==0.11.1
 Werkzeug==3.0.2
 wsproto==1.2.0
-zipp==3.18.1
+zipp==3.18.1
+streamlit-webrtc==0.47.6