import cv2
import time
import numpy as np
import onnxruntime
from scipy.special import softmax

# 加载ONNX模型
session = onnxruntime.InferenceSession("model/detect/best.onnx")
# 摄像头索引号，通常为0表示第一个摄像头
camera_index = 0

# 打开摄像头
cap = cv2.VideoCapture(camera_index, cv2.CAP_DSHOW)
# 设置分辨率
cap.set(cv2.CAP_PROP_FRAME_WIDTH, 3840)  # 宽度
cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 2160)  # 高度
# 检查摄像头是否成功打开
if not cap.isOpened():
    print("无法打开摄像头")
    exit()

width = cap.get(cv2.CAP_PROP_FRAME_WIDTH)
height = cap.get(cv2.CAP_PROP_FRAME_HEIGHT)
print("摄像头分辨率:", width, "x", height)



# 从res.json中读取类别
# with open("res1-2.json", "r") as f:
#     classes = eval(f.read())



# 目标图像尺寸
target_width = 1024
target_height = 768

# 计时器
start_time = time.time()

# 循环读取摄像头画面
while True:
    ret, frame = cap.read()

    if not ret:
        print("无法读取摄像头画面")
        break

    # 1920*1080的图像，中心裁剪640*480的区域
    cropped_frame = frame[int(height / 2 - target_height / 2):int(height / 2 + target_height / 2),
                    int(width / 2 - target_width / 2):int(width / 2 + target_width / 2)]
    # 调整图像尺寸
    resized_frame = cv2.resize(cropped_frame, (target_width, target_height))

    # 获取当前时间
    current_time = time.time()

    #如果距离上一次保存已经过去1秒，则保存当前画面
    # if current_time - start_time >= 3.0:
    #     # 生成保存文件名，以当前时间命名
    #     save_name = time.strftime("%Y%m%d%H%M%S", time.localtime()) + ".jpg"
    #     # 保存调整尺寸后的图片
    #     cv2.imwrite(save_path + save_name, frame)
    #     print("保存图片:", save_name)
    #     # 重置计时器
    #     start_time = time.time()

    # 预处理
    blob = cv2.dnn.blobFromImage(resized_frame, 1 / 255.0, (640, 640), swapRB=True, crop=False)

    # 模型推理
    outputs = session.run(None, {session.get_inputs()[0].name: blob})

    output = np.transpose(np.squeeze(outputs[0]));
    rows = output.shape[0]
    boxes = []
    scores = []
    class_ids = []
    for i in range(rows):
        classes_scores = output[i][4:]
        max_score = np.amax(classes_scores)
        if max_score > 0.5:
            classid = np.argmax(classes_scores)
            scores.append(max_score)
            class_ids.append(classid)

    print(class_ids)
    print(scores)
    # # 输出前十个类别
    # print("Top 5 Classes:")
    # for i in top_ten_classes:
    #     print(f"{classes[i]}: {probabilities[0][i]}")

    # 显示画面
    cv2.imshow("Camera", resized_frame)

    # 检测按键，如果按下q键则退出循环
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

# 关闭摄像头
cap.release()

# 关闭所有窗口
cv2.destroyAllWindows()