Refactor into modular app structure

Split monolithic files into focused modules: - app/core: settings, logging, lifecycle - app/signaling: websocket server, ICE parser, message models - app/webrtc: peer session, video receiver, frame source - app/vision: pose landmarker wrapper, model config, pose types - app/exercises/dead_bug: detector, metrics, rules, state machine, types - app/rendering: skeleton renderer, status overlay, window display - app/audio: rep announcer - app/diagnostics: perf timer, crash handler - configs: environment-based settings - tests: unit tests for rules, state machine, ICE parser - run.py: entry point
2026-06-10 10:14:43 +08:00
parent 8b878cb9e5
commit 4485cbf702
44 changed files with 1230 additions and 648 deletions
@@ -0,0 +1,106 @@
+from __future__ import annotations
+
+import asyncio
+import os
+
+import cv2
+from aiortc.mediastreams import MediaStreamError
+from loguru import logger
+
+from app.audio.rep_announcer import RepAnnouncer
+from app.exercises.dead_bug.detector import DeadBugDetector
+from app.rendering.window_display import close_window, is_esc_pressed, show_frame
+from configs.default import (
+    EXTENSION_CONFIRM_FRAMES,
+    MODEL_PATH,
+    PREFER_GPU,
+    PROCESS_EVERY_N_FRAMES,
+    REP_ANNOUNCER_ENABLED,
+    REP_ANNOUNCER_RATE,
+    REP_ANNOUNCER_VOLUME,
+    RESET_CONFIRM_FRAMES,
+    VISIBILITY_THRESHOLD,
+)
+
+
+def _format_pose_debug(pose_result) -> str:
+    metrics = pose_result.metrics
+    if metrics is None:
+        return "metrics=None"
+    return (
+        f"side={pose_result.side}, standard={pose_result.is_standard}, "
+        f"angles(le={metrics.left_elbow_angle:.1f}, re={metrics.right_elbow_angle:.1f}, "
+        f"lk={metrics.left_knee_angle:.1f}, rk={metrics.right_knee_angle:.1f}), "
+        f"extended(la={metrics.left_arm_extended}, ra={metrics.right_arm_extended}, "
+        f"ll={metrics.left_leg_extended}, rl={metrics.right_leg_extended})"
+    )
+
+
+class VideoReceiver:
+    def __init__(self, track) -> None:
+        self._track = track
+
+    async def run(self) -> None:
+        logger.info("Start receiving video frames, process_every_n={}", PROCESS_EVERY_N_FRAMES)
+
+        frame_count = 0
+        processed_count = 0
+        detector = DeadBugDetector(
+            model_path=MODEL_PATH,
+            visibility_threshold=VISIBILITY_THRESHOLD,
+            extension_confirm_frames=EXTENSION_CONFIRM_FRAMES,
+            reset_confirm_frames=RESET_CONFIRM_FRAMES,
+            prefer_gpu=PREFER_GPU,
+        )
+        announcer = RepAnnouncer(
+            enabled=REP_ANNOUNCER_ENABLED,
+            rate=REP_ANNOUNCER_RATE,
+            volume=REP_ANNOUNCER_VOLUME,
+        )
+        last_announced_rep = 0
+        last_pose_result = None
+        last_annotated = None
+
+        try:
+            while True:
+                frame = await self._track.recv()
+                frame_count += 1
+                raw_img = frame.to_ndarray(format="bgr24")
+                timestamp_ms = int(frame.time * 1000) if frame.time is not None else frame_count * 33
+
+                if frame_count % PROCESS_EVERY_N_FRAMES == 0 or last_pose_result is None:
+                    processed_count += 1
+                    last_annotated, last_pose_result = detector.process_frame(raw_img, timestamp_ms)
+                    if last_pose_result.rep_count > last_announced_rep:
+                        last_announced_rep = last_pose_result.rep_count
+                        announcer.announce_count(last_announced_rep)
+
+                display_img = last_annotated if last_annotated is not None else raw_img
+                show_frame(display_img)
+
+                if frame_count % 100 == 0:
+                    logger.info(
+                        "Received {} frames, processed={}, raw_shape={}, reps={}, phase={}, feedback={}, {}",
+                        frame_count,
+                        processed_count,
+                        raw_img.shape,
+                        last_pose_result.rep_count if last_pose_result is not None else 0,
+                        last_pose_result.phase.value if last_pose_result is not None else "none",
+                        " | ".join(last_pose_result.feedback) if last_pose_result is not None else "",
+                        _format_pose_debug(last_pose_result) if last_pose_result is not None else "metrics=None",
+                    )
+
+                if is_esc_pressed():
+                    logger.info("ESC pressed, closing display")
+                    break
+
+        except asyncio.CancelledError:
+            logger.info("Video receive task cancelled")
+        except MediaStreamError:
+            logger.info("Video track ended")
+        except Exception as e:
+            logger.exception(f"Video receive error: {e!r}")
+        finally:
+            announcer.close()
+            detector.close()
+            close_window()