Python+OpenCV实现用MediaPipe手势数字识别一键打开下载夹里的图片

py -3.10 -m venv venv
venv\Scripts\activate
pip install mediapipe opencv-python pillow numpy

https://storage.googleapis.com/mediapipe-models/hand_landmarker/hand_landmarker/float16/latest/hand_landmarker.task

import mediapipe as mp
from mediapipe.tasks import python as mp_python
from mediapipe.tasks.python import vision

options = vision.HandLandmarkerOptions(
    base_options=mp_python.BaseOptions(model_asset_path="models/hand_landmarker.task"),
    running_mode=vision.RunningMode.LIVE_STREAM,   # 摄像头流式推理
    num_hands=1,
    min_hand_detection_confidence=0.5,
    min_hand_presence_confidence=0.5,
    min_tracking_confidence=0.5,
    result_callback=on_result,                     # 异步结果回调
)

class LatestResult:
    def __init__(self):
        self.result = None

latest = LatestResult()

def on_result(result, output_image, timestamp_ms):
    latest.result = result

start_ns = time.perf_counter_ns()
# ...
rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
mp_image = mp.Image(image_format=mp.ImageFormat.SRGB, data=rgb)
ts_ms = (time.perf_counter_ns() - start_ns) // 1_000_000
landmarker.detect_async(mp_image, ts_ms)

HAND_CONNECTIONS = (
    (0,1),(1,2),(2,3),(3,4),
    (0,5),(5,6),(6,7),(7,8),
    (5,9),(9,10),(10,11),(11,12),
    (9,13),(13,14),(14,15),(15,16),
    (13,17),(0,17),(17,18),(18,19),(19,20),
)

FINGER_TIPS = (4, 8, 12, 16, 20)
FINGER_PIPS = (3, 6, 10, 14, 18)

def count_extended_fingers(landmarks) -> int:
    wrist = landmarks[0]
    def d(a, b):
        return math.hypot(a.x - b.x, a.y - b.y)
    n = 0
    for tip, pip in zip(FINGER_TIPS, FINGER_PIPS):
        if d(landmarks[tip], wrist) > d(landmarks[pip], wrist) * 1.1:
            n += 1
    return n

history = deque(maxlen=10)
history.append(cur_n)
stable = history[0] if len(history) == history.maxlen and all(v == history[0] for v in history) else None

if stable == 0:
    armed = True
elif stable is not None and 1 <= stable <= 5 and armed:
    open_nth_png(stable)
    armed = False
    history.clear()  # 清空缓冲，防止同一稳定段被重复识别

DOWNLOADS_DIR = Path.home() / "Downloads"

def open_nth_png(n: int):
    files = sorted(
        DOWNLOADS_DIR.glob("*.png"),
        key=lambda p: p.stat().st_mtime,
        reverse=True,
    )
    if len(files) < n:
        return False, f"Only {len(files)} PNG(s), need #{n}"
    target = files[n - 1]
    os.startfile(str(target))    # Windows 专属：相当于双击
    return True, f"Opened #{n}: {target.name}"

raw = result.handedness[hand_idx][0].category_name
label = "Right" if raw == "Left" else "Left"

with vision.HandLandmarker.create_from_options(options) as landmarker:
    while True:
        ok, frame = cap.read()
        if not ok:
            break
        frame = cv2.flip(frame, 1)
        rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
        mp_image = mp.Image(image_format=mp.ImageFormat.SRGB, data=rgb)
        ts_ms = (time.perf_counter_ns() - start_ns) // 1_000_000
        landmarker.detect_async(mp_image, ts_ms)

        res = latest.result
        cur_n = count_extended_fingers(res.hand_landmarks[0]) if (res and res.hand_landmarks) else 0
        history.append(cur_n)
        stable = history[0] if len(history) == history.maxlen \
                 and all(v == history[0] for v in history) else None

        if stable == 0:
            armed = True
        elif stable is not None and 1 <= stable <= 5 and armed:
            ok2, msg = open_nth_png(stable)
            armed = False
            history.clear()

        draw_landmarks(frame, res)
        cv2.imshow("Gesture -> Open PNG", frame)
        if cv2.waitKey(1) & 0xFF in (ord("q"), 27):
            break