Raspberry Pi3B + Pi Camera V2.1 で物体検出

$ sudo dmesg | grep imx

[    0.078042] /soc/csi@7e801000: Fixed dependency cycle(s) with /soc/i2c0mux/i2c@1/imx219@10
[    0.078123] /soc/i2c0mux/i2c@1/imx219@10: Fixed dependency cycle(s) with /soc/csi@7e801000

$ sudo apt udpate
$ sudo apt full-upgrade

sudo apt update
sudo apt install -y python3-opencv

#!/usr/bin/env python3
import time
import cv2
from picamera2 import Picamera2

def main():
    picam2 = Picamera2()
    config = picam2.create_preview_configuration(
        main={"size": (640, 480)},
        lores={"size": (160, 120)},
        display="main"
    )
    picam2.configure(config)
    picam2.start()

    # ② Haar Cascade の読み込み
    cascade_path = "./haarcascade_fullbody.xml"
#    cascade_path = "./haarcascade_fullbody.xml"
    cascade_path = "./haarcascade_eye.xml"
#    cascade_path = "./haarcascade_frontalface_default.xml"
    body_cascade = cv2.CascadeClassifier(cascade_path)

    cv2.namedWindow("Detection", cv2.WINDOW_NORMAL)

    try:
        while True:
            # ③ フレーム取得（NumPy 配列）
            frame = picam2.capture_array("main")      # BGR8888 の ndarray

            # ④ グレースケール変換＆検知
            gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)

            bodies = body_cascade.detectMultiScale(
                gray,
                scaleFactor=1.1,
                minNeighbors=3,
                minSize=(60, 60)
            )

            # ⑤ 検知結果を描画
            for (x, y, w, h) in bodies:
                cv2.rectangle(frame, (x, y), (x+w, y+h), (0,255,0), 2)

            # ⑥ ウィンドウ表示
            cv2.imshow("Detection", frame)
            # Esc キーで終了
            if cv2.waitKey(1) & 0xFF == 27:
                break

    finally:
        picam2.stop()
        cv2.destroyAllWindows()

if __name__ == "__main__":
    main()

pip3 install tflite-runtime

$ unzip 1.tflite
Archive:  1.tflite
 extracting: labelmap.txt            

#!/usr/bin/env python3
import cv2
import numpy as np
import tflite_runtime.interpreter as tflite
from picamera2 import Picamera2

# TFLite 用ヘルパー
class TFLiteModel:
    def __init__(self, model_path, label_path, threshold=0.5):
        self.interpreter = tflite.Interpreter(model_path=model_path)
        self.interpreter.allocate_tensors()
        self.input_details  = self.interpreter.get_input_details()
        self.output_details = self.interpreter.get_output_details()
        self.labels = [l.strip() for l in open(label_path)]
        self.threshold = threshold

    def detect(self, image):
        h, w, _ = image.shape
        inp_h, inp_w = self.input_details[0]['shape'][1:3]
        img = cv2.resize(image, (inp_w, inp_h))
        input_data = np.expand_dims(img, axis=0)
        self.interpreter.set_tensor(self.input_details[0]['index'], input_data)
        self.interpreter.invoke()

        boxes   = self.interpreter.get_tensor(self.output_details[0]['index'])[0]  # [N,4]
        classes = self.interpreter.get_tensor(self.output_details[1]['index'])[0]  # [N]
        scores  = self.interpreter.get_tensor(self.output_details[2]['index'])[0]  # [N]

        results = []
        for box, cls, score in zip(boxes, classes, scores):
            if score < self.threshold or int(cls) >= len(self.labels):
                continue
            ymin, xmin, ymax, xmax = box
            results.append({
                "label": self.labels[int(cls)],
                "score": float(score),
                "rect": (
                    int(xmin * w), int(ymin * h),
                    int((xmax - xmin) * w), int((ymax - ymin) * h)
                )
            })
        return results

def main():
    # カメラ
    picam2 = Picamera2()
    config = picam2.create_preview_configuration(
        main={"size": (640, 480), "format": "RGB888"},  # ←ここで３チャンネル指定
        display="main"
    )
    picam2.configure(config)
    picam2.start()

    # TFLite モデル読み込み
    model = TFLiteModel("detect.tflite", "labels.txt", threshold=0.6)

    cv2.namedWindow("TFLite Detection", cv2.WINDOW_NORMAL)
    try:
        while True:
            frame = picam2.capture_array("main")
            results = model.detect(frame)

            # 描画
            for r in results:
                x, y, w, h = r["rect"]
                cv2.rectangle(frame, (x,y), (x+w,y+h), (255,0,0), 2)
                cv2.putText(frame, f'{r["label"]}:{r["score"]:.2f}',
                            (x, y-10), cv2.FONT_HERSHEY_SIMPLEX, 1, (255,0,0), 2)

            cv2.imshow("TFLite Detection", frame)
            if cv2.waitKey(1) & 0xFF == 27:
                break

    finally:
        picam2.stop()
        cv2.destroyAllWindows()

if __name__ == "__main__":
    main()