Update emotion.py

emotion.py

@@ -1,4 +1,5 @@
 import time
+import threading
 from dataclasses import dataclass, field
 from typing import Optional, Dict, Tuple, List
 
@@ -9,31 +10,68 @@ from PIL import Image, ImageDraw
 from deepface import DeepFace
 
 
-# ===== 避免 css 未定義造成 NameError（不影響功能）=====
+# ===== 避免 css 未定義造成 NameError =====
 css = ""
 
 
 # =========================
-# 參數（可微調）
+# 參數（優先順：快）
 # =========================
-ANALYZE_EVERY_SEC = 0.8
-DOWNSAMPLE_W = 720
+ANALYZE_EVERY_SEC = 1.2     # ✅ 降頻率（原 0.8）
+DOWNSAMPLE_W = 480          # ✅ 降解析（原 720）
 EMA_ALPHA = 0.45
 CONF_TH = 40.0
 SWITCH_CONFIRM_SEC = 2.5
 IOU_TH = 0.25
-ZOOM_FACTOR = 1.5
+ZOOM_FACTOR = 1.25          # ✅ 稍微降 zoom（減少失真與負擔）
 
-DETECTOR_BACKEND = "mtcnn"  # 想快一點可改 "opencv"
+DETECTOR_BACKEND = "opencv" # ✅ 改快（原 mtcnn）
 ALIGN_FACE = True
 
-# ✅ 防止背景誤判為臉（重要）
-MIN_FACE_AREA_RATIO = 0.05  # 臉框面積佔整張畫面比例下限（3%~8%可調）
-MIN_DET_CONF = 0.90         # 偵測器信心值下限（若 backend 提供 confidence）
+# 防誤判（保留）
+MIN_FACE_AREA_RATIO = 0.05
+MIN_DET_CONF = 0.90
 
 
 # =========================
-# 情緒中文 + 分數映射（7 -> 5）
+# Warm-up（優先順：按 Record 不要等）
+# =========================
+_WARMED = False
+
+def _warmup_models():
+    global _WARMED
+    if _WARMED:
+        return
+    _WARMED = True
+    try:
+        dummy = np.zeros((224, 224, 3), dtype=np.uint8)
+        # detector warmup
+        try:
+            DeepFace.extract_faces(
+                img_path=dummy,
+                detector_backend=DETECTOR_BACKEND,
+                enforce_detection=False,
+                align=ALIGN_FACE,
+            )
+        except Exception:
+            pass
+        # emotion model warmup
+        try:
+            DeepFace.analyze(
+                img_path=dummy,
+                actions=["emotion"],
+                enforce_detection=False,
+            )
+        except Exception:
+            pass
+    except Exception:
+        pass
+
+threading.Thread(target=_warmup_models, daemon=True).start()
+
+
+# =========================
+# 情緒中文 + 分數映射
 # =========================
 EMO_ZH = {
     "angry": "生氣",
@@ -72,30 +110,23 @@ SCORE_CHOICES = [SCORE_LABELS[i] for i in [1, 2, 3, 4, 5]]
 # =========================
 @dataclass
 class AppState:
-    mode: str = "攝影機辨識"
-
     running: bool = True
     finished: bool = False
     last_analyze_t: float = 0.0
 
-    # tracking（用於連續幀追蹤）
     track_bbox: Optional[Tuple[int, int, int, int]] = None
 
-    # smoothing
     ema_scores: Optional[np.ndarray] = None
     emo_keys: List[str] = field(default_factory=list)
 
-    # stable/candidate
     stable_emo: str = "unknown"
     candidate_emo: Optional[str] = None
     candidate_since: Optional[float] = None
 
-    # final
     final_emo: Optional[str] = None
     final_conf: Optional[float] = None
     final_score: int = 3
 
-    # ✅ 新增：最近有沒有臉（避免節流期間框顏色亂跳）
     last_face_t: float = 0.0
     face_in_box: bool = False
 
@@ -127,33 +158,12 @@ def _result_card_html(emo_key: str, conf: Optional[float]) -> str:
     score = EMO_TO_SCORE.get(emo_key, 3)
     conf_txt = "" if conf is None else f"{conf:.1f}%"
     score_desc = SCORE_LABELS[score].split("（")[1].rstrip("）")
-
     return f"""
-    <div style="
-        border-radius:16px;
-        padding:20px;
-        border:1px solid #E6E6E6;
-        background:#FFFFFF;
-        color:#000000;
-    ">
-      <div style="font-size:14px;color:#000000;margin-bottom:8px;">
-        辨識結果
-      </div>
-      <div style="
-        font-size:42px;
-        font-weight:800;
-        line-height:1.1;
-        margin-bottom:14px;
-        color:#000000;
-      ">
-        {zh}
-      </div>
-      <div style="font-size:16px;color:#000000;margin-bottom:6px;">
-        心情分數：{score}（{score_desc}）
-      </div>
-      <div style="font-size:14px;color:#000000;">
-        信心值：{conf_txt}
-      </div>
+    <div style="border-radius:16px;padding:20px;border:1px solid #E6E6E6;background:#FFFFFF;color:#000000;">
+      <div style="font-size:14px;color:#000000;margin-bottom:8px;">辨識結果</div>
+      <div style="font-size:42px;font-weight:800;line-height:1.1;margin-bottom:14px;color:#000000;">{zh}</div>
+      <div style="font-size:16px;color:#000000;margin-bottom:6px;">心情分數：{score}（{score_desc}）</div>
+      <div style="font-size:14px;color:#000000;">信心值：{conf_txt}</div>
     </div>
     """
 
@@ -186,12 +196,11 @@ def _zoom_center(frame: np.ndarray, zoom: float) -> np.ndarray:
     new_h = int(h / zoom)
     start_x = (w - new_w) // 2
     start_y = (h - new_h) // 2
-    cropped = frame[start_y : start_y + new_h, start_x : start_x + new_w]
-    return cropped
+    return frame[start_y:start_y + new_h, start_x:start_x + new_w]
 
 
 # =========================
-# ��導框
+# 引導框（永遠畫）
 # =========================
 GUIDE_W_RATIO = 0.55
 GUIDE_H_RATIO = 0.70
@@ -217,9 +226,6 @@ def _bbox_center_in_box(bbox_xywh: Tuple[float, float, float, float], box_xyxy:
     return (x1 <= cx <= x2) and (y1 <= cy <= y2)
 
 def _draw_guide_overlay(frame_rgb: np.ndarray, detected_in_box: bool) -> np.ndarray:
-    if frame_rgb is None:
-        return frame_rgb
-
     img = Image.fromarray(frame_rgb)
     draw = ImageDraw.Draw(img)
 
@@ -228,7 +234,6 @@ def _draw_guide_overlay(frame_rgb: np.ndarray, detected_in_box: bool) -> np.ndar
 
     h, w = frame_rgb.shape[:2]
     thickness = max(3, int(min(w, h) * 0.008))
-
     for t in range(thickness):
         draw.rectangle([x1 - t, y1 - t, x2 + t, y2 + t], outline=color)
 
@@ -259,13 +264,9 @@ def _iou(a: Tuple[int, int, int, int], b: Tuple[int, int, int, int]) -> float:
     union = aw * ah + bw * bh - inter
     return inter / max(1e-6, union)
 
-def _pick_face_by_tracking(
-    faces: List[Dict],
-    prev_bbox: Optional[Tuple[int, int, int, int]]
-) -> Tuple[Optional[Dict], Optional[Tuple[int, int, int, int]]]:
+def _pick_face_by_tracking(faces: List[Dict], prev_bbox: Optional[Tuple[int, int, int, int]]):
     if not faces:
         return None, None
-
     cand = []
     for f in faces:
         fa = f.get("facial_area", {}) or {}
@@ -312,21 +313,12 @@ def _reset_state_for_camera(st: AppState) -> AppState:
     return st
 
 
-# =========================
-# ✅ 判斷這個 bbox 是否真的像「人臉」
-# =========================
-def _is_real_face(
-    pick: Dict,
-    bbox_full_xywh: Tuple[float, float, float, float],
-    frame_hw: Tuple[int, int],
-) -> bool:
+def _is_real_face(pick: Dict, bbox_full_xywh: Tuple[float, float, float, float], frame_hw: Tuple[int, int]) -> bool:
     fh, fw = frame_hw
     x, y, w, h = bbox_full_xywh
-
     area_ratio = (w * h) / max(1.0, float(fw * fh))
     if area_ratio < MIN_FACE_AREA_RATIO:
         return False
-
     det_conf = pick.get("confidence", None)
     if det_conf is not None:
         try:
@@ -334,31 +326,29 @@ def _is_real_face(
                 return False
         except Exception:
             pass
-
     return True
 
 
 # =========================
-# Camera streaming
+# Stream（優先順）
 # =========================
 def on_stream(frame_rgb: np.ndarray, st: AppState):
-    # 已結束/停止
     if st.finished or (not st.running):
         cam_visible = False
         if st.final_emo:
-            result_html = _result_card_html(st.final_emo, st.final_conf)
-            mood_val = _score_radio_value(EMO_TO_SCORE.get(st.final_emo, 3))
-            hint_html = _camera_hint_done()
-        else:
-            result_html = _hint_html("尚未完成辨識。")
-            mood_val = _score_radio_value(3)
-            hint_html = _camera_hint_stopped()
-
+            return (
+                gr.update(visible=cam_visible),
+                gr.update(value=_result_card_html(st.final_emo, st.final_conf)),
+                gr.update(value=_score_radio_value(EMO_TO_SCORE.get(st.final_emo, 3))),
+                gr.update(value=_camera_hint_done()),
+                st,
+                gr.update(value="重新攝影機辨識"),
+            )
         return (
             gr.update(visible=cam_visible),
-            gr.update(value=result_html),
-            gr.update(value=mood_val),
-            gr.update(value=hint_html),
+            gr.update(value=_hint_html("尚未完成辨識。")),
+            gr.update(value=_score_radio_value(3)),
+            gr.update(value=_camera_hint_stopped()),
             st,
             gr.update(value="重新攝影機辨識"),
         )
@@ -374,29 +364,27 @@ def on_stream(frame_rgb: np.ndarray, st: AppState):
         )
 
     frame_rgb = _zoom_center(frame_rgb, ZOOM_FACTOR)
-
     now = time.time()
 
-    # 節流：沿用最近一次臉狀態（避免顏色亂跳）
-    detected_in_box_recent = st.face_in_box and (now - st.last_face_t) < 1.0
+    # ✅ 先畫框（永遠立即有框），不等 detector
+    detected_recent = st.face_in_box and (now - st.last_face_t) < 1.0
+    annotated_fast = _draw_guide_overlay(frame_rgb, detected_recent)
 
+    # ✅ 如果還沒到分析時間，直接回傳畫框（超順）
     if now - st.last_analyze_t < ANALYZE_EVERY_SEC:
-        annotated = _draw_guide_overlay(frame_rgb, detected_in_box_recent)
-        result_html = _hint_html("已偵測到人臉，開始辨識情緒中…" if detected_in_box_recent else "未偵測到人臉，請把臉放進框內")
         return (
-            gr.update(visible=True, value=annotated),
-            gr.update(value=result_html),
+            gr.update(visible=True, value=annotated_fast),
+            gr.update(value=_hint_html("已偵測到人臉，開始辨識情緒中…" if detected_recent else "未偵測到人臉，請把臉放進框內")),
             gr.update(value=_score_radio_value(3)),
             gr.update(value=_camera_hint_before_start()),
             st,
             gr.update(value="重新攝影機辨識"),
         )
 
+    # 到時間���做一次重工作（低頻）
     st.last_analyze_t = now
-
     small = _downsample_rgb(frame_rgb, DOWNSAMPLE_W)
 
-    # ✅ extract_faces 變嚴格：沒臉就當 faces=[]
     try:
         faces = DeepFace.extract_faces(
             img_path=small,
@@ -409,13 +397,11 @@ def on_stream(frame_rgb: np.ndarray, st: AppState):
 
     pick, bbox = _pick_face_by_tracking(faces, st.track_bbox)
 
-    # (A) 完全沒臉
     if pick is None or bbox is None or pick.get("face") is None:
         st.track_bbox = None
         st.face_in_box = False
         st.last_face_t = now
         st.stable_emo = "unknown"
-
         annotated = _draw_guide_overlay(frame_rgb, False)
         return (
             gr.update(visible=True, value=annotated),
@@ -426,22 +412,17 @@ def on_stream(frame_rgb: np.ndarray, st: AppState):
             gr.update(value="重新攝影機辨識"),
         )
 
-    # bbox small -> frame
     sh, sw = small.shape[:2]
     fh, fw = frame_rgb.shape[:2]
     sx = fw / max(1, sw)
     sy = fh / max(1, sh)
-
     x, y, w, h = bbox
     bbox_full = (x * sx, y * sy, w * sx, h * sy)
 
-    # ✅ 防背景誤判：不夠大/信心不足 -> 視為沒臉
     if not _is_real_face(pick, bbox_full, (fh, fw)):
         st.track_bbox = None
         st.face_in_box = False
         st.last_face_t = now
-        st.stable_emo = "unknown"
-
         annotated = _draw_guide_overlay(frame_rgb, False)
         return (
             gr.update(visible=True, value=annotated),
@@ -453,14 +434,13 @@ def on_stream(frame_rgb: np.ndarray, st: AppState):
         )
 
     guide = _guide_box(frame_rgb)
-    detected_in_box = _bbox_center_in_box(bbox_full, guide)
+    in_box = _bbox_center_in_box(bbox_full, guide)
 
     st.last_face_t = now
-    st.face_in_box = detected_in_box
+    st.face_in_box = in_box
     st.track_bbox = bbox
 
-    # (B) 有臉但不在框內
-    if not detected_in_box:
+    if not in_box:
         annotated = _draw_guide_overlay(frame_rgb, False)
         return (
             gr.update(visible=True, value=annotated),
@@ -471,12 +451,11 @@ def on_stream(frame_rgb: np.ndarray, st: AppState):
             gr.update(value="重新攝影機辨識"),
         )
 
-    # (C) 有臉且在框內：綠框，開始 analyze
+    # 只有臉在框內才分析情緒
     face = pick["face"]
     if face.dtype != np.uint8:
         face = np.clip(face * 255.0, 0, 255).astype(np.uint8)
 
-    # ✅ analyze 也嚴格：避免誤判一路走到底
     try:
         result = DeepFace.analyze(
             img_path=face,
@@ -484,12 +463,9 @@ def on_stream(frame_rgb: np.ndarray, st: AppState):
             enforce_detection=True,
         )
     except Exception:
-        st.track_bbox = None
+        annotated = _draw_guide_overlay(frame_rgb, False)
         st.face_in_box = False
         st.last_face_t = now
-        st.stable_emo = "unknown"
-
-        annotated = _draw_guide_overlay(frame_rgb, False)
         return (
             gr.update(visible=True, value=annotated),
             gr.update(value=_hint_html("未偵測到人臉，請把臉放進框內")),
@@ -502,7 +478,6 @@ def on_stream(frame_rgb: np.ndarray, st: AppState):
     if isinstance(result, list):
         result = result[0]
     emo_dict = result.get("emotion", None)
-
     if not isinstance(emo_dict, dict) or len(emo_dict) == 0:
         annotated = _draw_guide_overlay(frame_rgb, True)
         return (
@@ -541,7 +516,6 @@ def on_stream(frame_rgb: np.ndarray, st: AppState):
     stable_key = st.emo_keys[stable_idx]
     st.stable_emo = stable_key
 
-    # 延遲確認
     if st.candidate_emo != stable_key:
         st.candidate_emo = stable_key
         st.candidate_since = now
@@ -558,10 +532,8 @@ def on_stream(frame_rgb: np.ndarray, st: AppState):
         st.final_emo = stable_key
         st.final_conf = top_conf
         st.final_score = EMO_TO_SCORE.get(stable_key, 3)
-
         st.running = False
         st.finished = True
-
         return (
             gr.update(visible=False),
             gr.update(value=_result_card_html(stable_key, top_conf)),
@@ -588,7 +560,7 @@ def on_stream(frame_rgb: np.ndarray, st: AppState):
 def on_restart(st: AppState):
     st = _reset_state_for_camera(st)
     return (
-        gr.update(visible=True, value=None),  # ✅ 清掉上一張畫面（你問的重點）
+        gr.update(visible=True, value=None),  # ✅ 清掉上一張畫面
         gr.update(value=_hint_html("等待影像…（尚未開始串流）")),
         gr.update(value=_score_radio_value(3)),
         gr.update(value=_camera_hint_before_start()),
@@ -599,19 +571,17 @@ def on_restart(st: AppState):
 def on_stop(st: AppState):
     st.running = False
     st.finished = True
-
     if st.final_emo is None:
         return (
-            gr.update(visible=False, value=None),  # ✅ 順便清空畫面（可選，但我先幫你加上）
+            gr.update(visible=False, value=None),
             gr.update(value=_hint_html("已停止辨識。")),
             gr.update(value=_score_radio_value(3)),
             gr.update(value=_camera_hint_stopped()),
             st,
             gr.update(value="重新攝影機辨識"),
         )
-
     return (
-        gr.update(visible=False, value=None),  # ✅ 順便清空
+        gr.update(visible=False, value=None),
         gr.update(value=_result_card_html(st.final_emo, st.final_conf)),
         gr.update(value=_score_radio_value(EMO_TO_SCORE.get(st.final_emo, 3))),
         gr.update(value=_camera_hint_done()),