fixing camera n1235

requirements.txt

@@ -1,6 +1,7 @@
 altair
 pandas
 streamlit
+streamlit-webrtc
 torch
 torchvision
 opencv-python-headless

src/streamlit_app.py

@@ -14,6 +14,9 @@ import os
 import logging
 import warnings
 import requests
+import threading
+import av
+from streamlit_webrtc import webrtc_streamer, VideoProcessorBase
 
 # Suppress WebRTC/asyncio warnings and errors
 logging.getLogger('aioice').setLevel(logging.CRITICAL)
@@ -320,6 +323,32 @@ def create_confidence_chart(detections):
     
     return fig
 
+class RealtimeVideoProcessor(VideoProcessorBase):
+    def __init__(self, model, conf_threshold):
+        self.model = model
+        self.conf_threshold = conf_threshold
+        self.last_detections = None
+        self.lock = threading.Lock()
+
+    def recv(self, frame):
+        img_bgr = frame.to_ndarray(format="bgr24")
+        img_rgb = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2RGB)
+
+        if self.model is None:
+            return frame
+
+        self.model.conf = self.conf_threshold
+        results = self.model(img_rgb)
+
+        detections = results.pandas().xyxy[0]
+        relevant_detections = detections[detections['name'].isin(WHEELCHAIR_CLASSES.values())]
+
+        with self.lock:
+            self.last_detections = relevant_detections
+
+        rendered_img = results.render()[0]
+        return av.VideoFrame.from_ndarray(rendered_img, format="bgr24")
+
 # Main app
 def main():
     # Header
@@ -493,89 +522,45 @@ def main():
     
     with tab2:
         st.markdown("### 📹 Realtime Camera Detection")
-        st.caption("Allow camera access, then take a photo to run detection.")
-
-        camera_image = st.camera_input(
-            "Capture an image",
-            key="camera_input"
-        )
-
-        if camera_image is not None:
-            try:
-                image = Image.open(camera_image)
-            except Exception as e:
-                st.error(f"Could not read camera image: {e}")
-                image = None
-
-            if image is not None:
-                col1, col2 = st.columns(2)
-
-                with col1:
-                    st.markdown("#### 📸 Captured Image")
-                    st.image(image, width=500)
-
-                # Load model
-                if st.session_state.model is None:
-                    with st.spinner("🤖 Loading AI model..."):
-                        st.session_state.model = load_model(conf_threshold)
-
-                if st.session_state.model is not None:
-                    # Process image
-                    with st.spinner("🔍 Analyzing image for obstacles..."):
-                        start_time = time.time()
-                        rendered_img, detections, results = process_image(image, st.session_state.model, conf_threshold)
-                        processing_time = time.time() - start_time
-
-                    with col2:
-                        st.markdown("#### 🎯 Detection Results")
-                        if rendered_img is not None:
-                            st.image(rendered_img, width=500)
-
-                    # Generate navigation advice
-                    advice, safety_level = get_navigation_advice(detections, image.width)
+        st.caption("Start the camera to run live detection on each frame.")
 
-                    # Update safety status
-                    st.markdown("### 🛡️ Safety Status")
-                    if "CRITICAL" in safety_level:
-                        st.markdown(f'<div class="safety-alert"><strong>{safety_level}</strong></div>', unsafe_allow_html=True)
-                    elif "CAUTION" in safety_level:
-                        st.markdown(f'<div style="background: #fff3cd; color: #856404; padding: 1rem; border-radius: 8px; border-left: 4px solid #ffc107;"><strong>{safety_level}</strong></div>', unsafe_allow_html=True)
-                    else:
-                        st.markdown(f'<div class="success-alert"><strong>{safety_level}</strong></div>', unsafe_allow_html=True)
-
-                    # Display navigation advice
-                    st.markdown("### 🧭 Navigation Advice")
-                    for advice_text in advice:
-                        st.markdown(f'<div class="detection-box">{advice_text}</div>', unsafe_allow_html=True)
-
-                    # Performance metrics
-                    st.markdown("### ⚡ Performance Metrics")
-                    perf_col1, perf_col2, perf_col3, perf_col4 = st.columns(4)
-
-                    with perf_col1:
-                        st.metric("⏱️ Processing Time", f"{processing_time:.2f}s")
-
-                    with perf_col2:
-                        fps = 1 / processing_time if processing_time > 0 else 0
-                        st.metric("🎬 Estimated FPS", f"{fps:.1f}")
+        if st.session_state.model is None:
+            with st.spinner("🤖 Loading AI model..."):
+                st.session_state.model = load_model(conf_threshold)
 
-                    with perf_col3:
-                        total_detections = len(detections) if detections is not None else 0
-                        st.metric("🔍 Objects Detected", total_detections)
+        webrtc_ctx = webrtc_streamer(
+            key="realtime_camera",
+            video_processor_factory=lambda: RealtimeVideoProcessor(st.session_state.model, conf_threshold),
+            media_stream_constraints={"video": True, "audio": False},
+            async_processing=True
+        )
 
-                    with perf_col4:
-                        relevant_count = len(detections) if detections is not None else 0
-                        st.metric("🎯 Relevant Objects", relevant_count)
+        realtime_detections = None
+        if webrtc_ctx.video_processor:
+            with webrtc_ctx.video_processor.lock:
+                realtime_detections = webrtc_ctx.video_processor.last_detections
 
-                    # Detailed detection results
-                    if detections is not None and len(detections) > 0:
-                        st.markdown("### 📋 Detailed Detection Results")
+        if realtime_detections is not None:
+            st.markdown("### 🛡️ Safety Status")
+            advice, safety_level = get_navigation_advice(realtime_detections, image_width=640)
 
-                        display_df = detections[['name', 'confidence', 'xmin', 'ymin', 'xmax', 'ymax']].copy()
-                        display_df['confidence'] = display_df['confidence'].apply(lambda x: f"{x:.1%}")
-                        display_df.columns = ['Object', 'Confidence', 'X Min', 'Y Min', 'X Max', 'Y Max']
+            if "CRITICAL" in safety_level:
+                st.markdown(f'<div class="safety-alert"><strong>{safety_level}</strong></div>', unsafe_allow_html=True)
+            elif "CAUTION" in safety_level:
+                st.markdown(f'<div style="background: #fff3cd; color: #856404; padding: 1rem; border-radius: 8px; border-left: 4px solid #ffc107;"><strong>{safety_level}</strong></div>', unsafe_allow_html=True)
+            else:
+                st.markdown(f'<div class="success-alert"><strong>{safety_level}</strong></div>', unsafe_allow_html=True)
 
-                        st.dataframe(display_df, use_container_width=True)
+            st.markdown("### 🧭 Navigation Advice")
+            for advice_text in advice:
+                st.markdown(f'<div class="detection-box">{advice_text}</div>', unsafe_allow_html=True)
+
+            if len(realtime_detections) > 0:
+                st.markdown("### 📋 Latest Detections")
+                display_df = realtime_detections[['name', 'confidence', 'xmin', 'ymin', 'xmax', 'ymax']].copy()
+                display_df['confidence'] = display_df['confidence'].apply(lambda x: f"{x:.1%}")
+                display_df.columns = ['Object', 'Confidence', 'X Min', 'Y Min', 'X Max', 'Y Max']
+                st.dataframe(display_df, use_container_width=True)
 
     with tab3:
         st.markdown("### 📊 Detection Analytics")