Update app.py

app.py

@@ -2,6 +2,7 @@
 Streamlit Web Interface – Traffic Detection
 Supports: YOLOv11 (tracking) | SSD MobileNetV3
 GPU forced automatically.
+Upload and processing are separated to avoid 403 errors.
 """
 
 import streamlit as st
@@ -16,13 +17,20 @@ from ultralytics import YOLO
 from datetime import datetime
 from utils.yolo_tracker import normalize_class_name
 
+# ── Session state init ─────────────────────────────────────────────────────────
 if 'global_unique_ids' not in st.session_state:
     st.session_state.global_unique_ids = {}
-
-def refine_class_by_shape(cls_name: str, x1: int, y1: int, x2: int, y2: int) -> str:
+if 'video_ready' not in st.session_state:
+    st.session_state.video_ready = False
+if 'tmp_path' not in st.session_state:
+    st.session_state.tmp_path = None
+if 'processing_done' not in st.session_state:
+    st.session_state.processing_done = False
+
+def refine_class_by_shape(cls_name, x1, y1, x2, y2, ratio=1.2):
     w = max(1, x2 - x1)
     h = max(1, y2 - y1)
-    if cls_name == "car" and (h / w) > 1.2:
+    if cls_name == "car" and (h / w) > ratio:
         return "person"
     return cls_name
 
@@ -60,13 +68,13 @@ if model_choice == "YOLOv11 + ByteTrack (Base)":
 elif model_choice == "YOLOv11 + ByteTrack (Fine-tuned)":
     selected_yolo_path = YOLO_FINETUNED_PATH
     if os.path.exists(selected_yolo_path):
-        st.sidebar.success("Weights: `models/best.pt` ")
+        st.sidebar.success("Weights: `models/best.pt` ✅")
     else:
         st.sidebar.error("`models/best.pt` not found.")
 else:
     selected_yolo_path = None
 
-selected_classes = st.sidebar.multiselect(
+selected_classes     = st.sidebar.multiselect(
     "Classes to detect", TRAFFIC_CLASSES,
     default=['car', 'truck', 'bus', 'person', 'traffic light']
 )
@@ -76,33 +84,77 @@ min_track_hits       = st.sidebar.number_input("Min track hits (YOLO only)", min
 shape_refine         = st.sidebar.checkbox("Reclassify tall 'car' as person", value=True)
 person_car_ratio     = st.sidebar.slider("Person/Car ratio (H/W)", 0.8, 2.0, 1.2, 0.05)
 
+if not selected_classes:
+    st.warning("Select at least one class in the sidebar.")
+    st.stop()
+
+# ══════════════════════════════════════════════════════════════════════════════
+# ÉTAPE 1 — UPLOAD
+# ══════════════════════════════════════════════════════════════════════════════
+st.subheader("📤 Step 1 — Upload your video")
+
 uploaded_file = st.file_uploader(
-    "📤 Upload a video (mp4, avi, mov)",
+    "Choose a video file (mp4, avi, mov)",
     type=["mp4", "avi", "mov"]
 )
 
-# ── Main ───────────────────────────────────────────────────────────────────────
-if not selected_classes:
-    st.warning("Select at least one class in the sidebar.")
-
-elif uploaded_file:
-    tfile = tempfile.NamedTemporaryFile(delete=False, suffix='.mp4')
-    tfile.write(uploaded_file.read())
-    tfile.flush()
-
-    col_video, col_stats = st.columns([2, 1])
-    with col_video:
-        st.subheader("📹 Annotated video")
-        frame_display  = st.empty()
-        status_display = st.empty()
-    with col_stats:
-        st.subheader("🔢 Unique object counters")
-        counter_display = st.empty()
-        progress_bar    = st.progress(0)
-
-    if st.button("▶️ Start Detection", type="primary"):
+if uploaded_file is not None:
+    # Sauvegarde du fichier en session pour ne pas le reperdre
+    if not st.session_state.video_ready:
+        with st.spinner("Saving video..."):
+            tfile = tempfile.NamedTemporaryFile(delete=False, suffix='.mp4')
+            tfile.write(uploaded_file.read())
+            tfile.flush()
+            tfile.close()
+            st.session_state.tmp_path = tfile.name
+            st.session_state.video_ready = True
+            st.session_state.processing_done = False
+            st.session_state.global_unique_ids = {}
+
+    st.success(f"✅ Video ready : **{uploaded_file.name}** "
+               f"({uploaded_file.size / 1e6:.1f} MB)")
+
+    # Aperçu info vidéo
+    cap_info = cv.VideoCapture(st.session_state.tmp_path)
+    fps_info   = cap_info.get(cv.CAP_PROP_FPS)
+    w_info     = int(cap_info.get(cv.CAP_PROP_FRAME_WIDTH))
+    h_info     = int(cap_info.get(cv.CAP_PROP_FRAME_HEIGHT))
+    total_info = int(cap_info.get(cv.CAP_PROP_FRAME_COUNT))
+    cap_info.release()
+
+    col1, col2, col3, col4 = st.columns(4)
+    col1.metric("Resolution", f"{w_info}x{h_info}")
+    col2.metric("FPS", f"{fps_info:.1f}")
+    col3.metric("Frames", total_info)
+    col4.metric("Duration", f"{total_info/fps_info:.1f}s" if fps_info > 0 else "N/A")
+
+    st.divider()
+
+    # ══════════════════════════════════════════════════════════════════════════
+    # ÉTAPE 2 — TRAITEMENT
+    # ══════════════════════════════════════════════════════════════════════════
+    st.subheader("▶️ Step 2 — Run Detection")
+
+    start = st.button("▶️ Start Detection", type="primary",
+                      disabled=st.session_state.processing_done)
+
+    if start:
+        st.session_state.processing_done = False
         st.session_state.global_unique_ids = {}
 
+        col_video, col_stats = st.columns([2, 1])
+        with col_video:
+            st.subheader("📹 Annotated video")
+            frame_display  = st.empty()
+            status_display = st.empty()
+        with col_stats:
+            st.subheader("🔢 Unique counters")
+            counter_display = st.empty()
+            progress_bar    = st.progress(0)
+
+        tmp_path = st.session_state.tmp_path
+        logs     = []
+
         # ── YOLOv11 ───────────────────────────────────────────────────────────
         if model_choice in ("YOLOv11 + ByteTrack (Base)",
                             "YOLOv11 + ByteTrack (Fine-tuned)"):
@@ -111,11 +163,10 @@ elif uploaded_file:
                 st.stop()
 
             model        = YOLO(selected_yolo_path)
-            cap          = cv.VideoCapture(tfile.name)
+            cap          = cv.VideoCapture(tmp_path)
             fps          = cap.get(cv.CAP_PROP_FPS) or 25
             total        = int(cap.get(cv.CAP_PROP_FRAME_COUNT))
             track_hits   = {}
-            logs         = []
             frame_idx    = 0
             log_path     = f"logs/log_{datetime.now().strftime('%Y%m%d_%H%M%S')}.csv"
             name_to_id   = {v: k for k, v in model.names.items()}
@@ -145,10 +196,8 @@ elif uploaded_file:
                         x1, y1, x2, y2 = map(int, box.xyxy[0])
 
                         if shape_refine:
-                            w = max(1, x2 - x1)
-                            h = max(1, y2 - y1)
-                            if cls_name == "car" and (h / w) > person_car_ratio:
-                                cls_name = "person"
+                            cls_name = refine_class_by_shape(
+                                cls_name, x1, y1, x2, y2, person_car_ratio)
 
                         area = max(0, x2-x1) * max(0, y2-y1)
                         if cls_name not in selected_classes: continue
@@ -176,7 +225,8 @@ elif uploaded_file:
 
                 if no_object:
                     cv.putText(annotated, "No selected object detected",
-                               (20, 50), cv.FONT_HERSHEY_SIMPLEX, 1.0, (0,0,255), 2)
+                               (20, 50), cv.FONT_HERSHEY_SIMPLEX,
+                               1.0, (0, 0, 255), 2)
                     status_display.warning("⚠️ No selected object in this frame")
                 else:
                     status_display.empty()
@@ -208,7 +258,7 @@ elif uploaded_file:
                                  'class', 'x1', 'y1', 'x2', 'y2', 'confidence'])
                 writer.writerows(logs)
 
-            st.success(f"Done — {frame_idx} frames analyzed")
+            st.success(f"✅ Done — {frame_idx} frames analyzed")
 
         # ── SSD ───────────────────────────────────────────────────────────────
         else:
@@ -241,15 +291,14 @@ elif uploaded_file:
             ssd_model.to(device)
             ssd_model.eval()
 
-            cap       = cv.VideoCapture(tfile.name)
+            cap       = cv.VideoCapture(tmp_path)
             fps       = cap.get(cv.CAP_PROP_FPS) or 25
             total     = int(cap.get(cv.CAP_PROP_FRAME_COUNT))
-            logs      = []
             frame_idx = 0
             log_path  = f"logs/ssd_log_{datetime.now().strftime('%Y%m%d_%H%M%S')}.csv"
             np.random.seed(42)
-            COLORS    = np.random.randint(0, 255, size=(len(COCO_CLASSES), 3),
-                                          dtype="uint8")
+            COLORS       = np.random.randint(0, 255, size=(len(COCO_CLASSES), 3),
+                                             dtype="uint8")
             seen_classes = set()
 
             while cap.isOpened():
@@ -279,10 +328,8 @@ elif uploaded_file:
 
                     x1, y1, x2, y2 = map(int, box)
                     if shape_refine:
-                        w = max(1, x2 - x1)
-                        h = max(1, y2 - y1)
-                        if cls_name == "car" and (h / w) > person_car_ratio:
-                            cls_name = "person"
+                        cls_name = refine_class_by_shape(
+                            cls_name, x1, y1, x2, y2, person_car_ratio)
 
                     area = max(0, x2-x1) * max(0, y2-y1)
                     if cls_name not in selected_classes: continue
@@ -332,7 +379,9 @@ elif uploaded_file:
                                  'x1','y1','x2','y2','confidence'])
                 writer.writerows(logs)
 
-            st.success(f"Done — {frame_idx} frames analyzed")
+            st.success(f"✅ Done — {frame_idx} frames analyzed")
+
+        st.session_state.processing_done = True
 
         # ── Final Stats ────────────────────────────────────────────────────────
         if logs:
@@ -371,7 +420,15 @@ elif uploaded_file:
                 mime="text/csv"
             )
 
-    os.unlink(tfile.name)
+            if st.button("🔄 Process another video"):
+                st.session_state.video_ready    = False
+                st.session_state.tmp_path       = None
+                st.session_state.processing_done = False
+                st.session_state.global_unique_ids = {}
+                st.rerun()
 
 else:
+    st.session_state.video_ready     = False
+    st.session_state.tmp_path        = None
+    st.session_state.processing_done = False
     st.info("👆 Upload a video to start.")