百度飞浆目标检测PPYOLOE模型在PC端、Jetson上的部署（python）

部署目标检测模型前，需要配置好paddlepaddle的环境：
开始使用_飞桨-源于产业实践的开源深度学习平台 (paddlepaddle.org.cn)

PC端和Jetson板卡端的部署方法相同，如下（直接放置部署和测试代码）：

import paddle.inference
import cv2
import numpy as np
import time
from paddle.inference import Config, PrecisionType, create_predictor
import os
import json
import argparse
from PIL import Image
import yaml
import globdef config_init(model_dir):model_file = glob.glob(model_dir + "/*.pdmodel")[0]params_file = glob.glob(model_dir + "/*.pdiparams")[0]config = Config()config.set_prog_file(model_file)config.set_params_file(params_file)config.switch_ir_optim()  # 开启IR优化config.enable_memory_optim()  # 启用内存优化config.enable_use_gpu(500, 0)  # 500是内存大小，0是GPU编号return create_predictor(config)  # 初始化一个预测器对象class PPYoloeModel:def __init__(self, model_dir):self.yaml_file = glob.glob(model_dir + "/*.yml")[0]self.predictor = config_init(model_dir)  # 初始化一个预测器对象self.mean = [0.0, 0.0, 0.0]self.std = [1.0, 1.0, 1.0]self.img_size = 640self.threshold = 0.6  # 置信度阈值def read_yaml(self):with open(self.yaml_file, 'r', encoding='utf-8') as infer_cfg:  # 使用with语句确保文件正确关闭yaml_reader = yaml.safe_load(infer_cfg)  # 直接从文件对象中读取label_list = yaml_reader['label_list']  # 获取分类# print(label_list)return label_listdef resize(self, img):if not isinstance(img, np.ndarray):raise TypeError('image type is not numpy.')im_shape = img.shape im_scale_x = float(self.img_size) / float(im_shape[1])  # 可以保持宽高比im_scale_y = float(self.img_size) / float(im_shape[0])img = cv2.resize(img, None, None, fx=im_scale_x, fy=im_scale_y)  # fx和fy是缩放因子（倍数）return imgdef normalize(self, img):img = img / 255.0  # 归一化到[0,1]mean = np.array(self.mean)[np.newaxis, np.newaxis, :]  # np.newaxis增加新的轴（维度）std = np.array(self.std)[np.newaxis, np.newaxis, :]img -= mean  img /= std return imgdef pre_process(self, img):img = self.resize(img)img = img[:, :, ::-1].astype('float32')  # bgr -> rgbimg = self.normalize(img)img = img.transpose((2, 0, 1))  # hwc -> chwreturn img[np.newaxis, :]def run_predict(self, img):input_names = self.predictor.get_input_names()for i, name in enumerate(input_names):input_tensor = self.predictor.get_input_handle(name)  # 获取张量句柄input_tensor.reshape(img[i].shape)  # 重塑输入张量input_tensor.copy_from_cpu(img[i].copy())  # 从CPU复制图像数据到输入张量self.predictor.run()results = []output_names = self.predictor.get_output_names()for i, name in enumerate(output_names):output_tensor = self.predictor.get_output_handle(name)output_data = output_tensor.copy_to_cpu()  # 获取检测结果results.append(output_data)return resultsdef get_result(self, img):result = []scale_factor = (np.array([self.img_size * 1.0 / img.shape[0], self.img_size * 1.0 / img.shape[1]]).reshape((1, 2)).astype(np.float32))img = self.pre_process(img)results = self.run_predict([img, scale_factor])for res in results[0]:score = res[1]if score > self.threshold:result.append(res)  # 只保留置信度大于阈值的目标（同时可能有多个目标）return result# result是一个列表，里面可能存有一个或多个目标的信息# 对于单个目标，第一个值是类别的id，第二个值是置信度，后四个值分别是xmin, ymin, xmax, ymaxdef draw_bbox_cv(self, img):  # 对于实时摄像头图片的后处理label_list = self.read_yaml()result = self.get_result(img)for res in result:cat_id, score, bbox = res[0], res[1], res[2:]  # cat_id表示类别的id；score是得分；bbox是列表，含有xmin, ymin, xmax, ymax四个元素bbox = [int(i) for i in bbox]  # 对坐标取整xmin, ymin, xmax, ymax = bboxcv2.rectangle(img, (xmin, ymin), (xmax, ymax), (255, 0, 255), 2)  # 2表示线宽print('category id is {}, bbox is {}'.format(cat_id, bbox))print(f"中心点：{(xmin + xmax) / 2},{(ymin + ymax) / 2}")try:label_id = label_list[int(cat_id)]  # 通过类别id查找对应类别名称# 在图像上打印类别   FONT_HERSHEY_SIMPLEX为字体类型cv2.putText(img, label_id, (int(xmin), int(ymin + 15)), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (255, 255, 0),2)# 在图像上打印得分cv2.putText(img, str(round(score, 2)), (int(xmin - 35), int(ymin + 15)), cv2.FONT_HERSHEY_SIMPLEX, 0.5,(0, 255, 0), 2)print(round(score, 2))except KeyError:pass# 调用示例
if __name__ == "__main__":ppyoloe_model = PPYoloeModel("这里输入模型所在文件夹的路径")  # 初始化一个ppyoloe目标检测模型对象cap = cv2.VideoCapture(0)  # 初始化摄像头while True:ret, image = cap.read()  # 获取图像if not ret:print("无法正确读取图像！")breakppyoloe_model.draw_bbox_cv(image)cv2.imshow("frame", image)if cv2.waitKey(1) & 0xFF == ord('q'):break