openvino +yolov5 自己训练模型并测试

Posted Arno1988

tags:

篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了openvino +yolov5 自己训练模型并测试相关的知识,希望对你有一定的参考价值。

https://github.com/ultralytics/yolov5/releases/tag/v6.1

conda create -n openvino_yolov6 python=3.7 -y
conda activate openvino_yolov6
pip install -r requirements.txt

export.py
export_onnx(model, im, file, 10, train, False, simplify) # opset 12 10

python export.py --weights yolov5n.pt --img 640 --batch 1

* 设置一个临时的环境变量
cd C:\\Program Files (x86)\\Intel\\openvino_2021.4.752\\bin
setupvars.bat

* 安装一些常用的模型优化器
cd C:\\Program Files (x86)\\Intel\\openvino_2021.4.752\\deployment_tools\\model_optimizer\\install_prerequisites
install_prerequisites.bat

* 模型转换
cd C:\\Program Files (x86)\\Intel\\openvino_2021.4.752\\deployment_tools\\model_optimizer
python mo_onnx.py --input_model H:\\tt\\yolov5-6.1\\yolov5n.onnx --output_dir E:\\yolo5n_IR\\

测试
python detect.py --weights yolov5n.pt --source data/images/bus.jpg

训练 yolov5 - CHHC - 博客园 (cnblogs.com)

 

python train.py --weights yolov5s.pt --data data/catdog.yaml --workers 1 --batch-size 8

训练测试
python detect.py --weights catdog.pt --source data/images/1.png
conda activate openvino_yolov6
python export.py --weights catdog.pt --img 640 --batch 1
python mo_onnx.py --input_model H:\\tt\\yolov5-6.1\\catdog.onnx --output_dir E:\\IR\\catdog

 

app.py

#!/usr/local/bin/python3
# encodin: utf-8
import cv2
import time
import os

from OpenVinoYoloV5Detector import OpenVinoYoloV5Detector

classes = []

classes_base = ['person', 'bicycle', 'car', 'motorcycle', 'airplane', 'bus', 'train', 'truck', 'boat', 'traffic light',
        'fire hydrant', 'stop sign', 'parking meter', 'bench', 'bird', 'cat', 'dog', 'horse', 'sheep', 'cow',
        'elephant', 'bear', 'zebra', 'giraffe', 'backpack', 'umbrella', 'handbag', 'tie', 'suitcase', 'frisbee',
        'skis', 'snowboard', 'sports ball', 'kite', 'baseball bat', 'baseball glove', 'skateboard', 'surfboard',
        'tennis racket', 'bottle', 'wine glass', 'cup', 'fork', 'knife', 'spoon', 'bowl', 'banana', 'apple',
        'sandwich', 'orange', 'broccoli', 'carrot', 'hot dog', 'pizza', 'donut', 'cake', 'chair', 'couch',
        'potted plant', 'bed', 'dining table', 'toilet', 'tv', 'laptop', 'mouse', 'remote', 'keyboard', 'cell phone',
        'microwave', 'oven', 'toaster', 'sink', 'refrigerator', 'book', 'clock', 'vase', 'scissors', 'teddy bear',
        'hair drier', 'toothbrush']  # class names

classes_catdog = ['cat', 'dog']  # class names

def get_milsecond():
    t = time.time()
    return (int(round(t * 1000)))


if __name__ == '__main__':
    url = '1.png'
    box_color = (0, 255, 0)

    # yolov5
    conf = 
        # "weight_file": "weights/yolov5n_openvino_model/yolov5n.xml",
        # "weight_file": "weights/yolov5s_openvino_model/yolov5s.xml",
        "weight_file": "weights/catdog_openvino_model/catdog.xml",
        "device": "CPU"
    
    classes = classes_catdog
    detector = OpenVinoYoloV5Detector(IN_conf=conf)

    # ssd
    # conf = 
    #     "model_xml": "./weights/ssdlite_mobilenet_v2/FP16/ssdlite_mobilenet_v2.xml",
    #     "model_bin": "./weights/ssdlite_mobilenet_v2/FP16/ssdlite_mobilenet_v2.bin",
    #     "device": "CPU"
    # 
    # rtscap.detector = OpenVinoSSDDetector(IN_conf=conf)

    cap = cv2.VideoCapture(url)

    while True:
        ret, frame = cap.read()
        if not ret or frame is None:
            break

        starttime = get_milsecond()
        detect_num, detect_data = detector.detect(frame)
        if len(detect_data):
            #cv2.imwrite('1.png', frame, [int(cv2.IMWRITE_JPEG_QUALITY), 95])

            for m in detect_data:
                classid= int(m.get('class_id'))
                classname = classes[classid]
                score = m.get('score')
                location = m.get('location')
                box_l, box_t = int(location.get('x1')), int(location.get('y1'))
                box_r, box_b = int(location.get('x2')), int(location.get('y2'))
                frame = cv2.rectangle(frame, (box_l, box_t), (box_r, box_b), box_color, 2)
                frame = cv2.putText(frame, classname + " " + str(score), (box_l, box_t + 15), cv2.FONT_HERSHEY_SIMPLEX, 0.4, (0, 0, 255), 2)

        endtime = get_milsecond();
        print('检测耗时:' + str(endtime - starttime))

        cv2.imshow('openvino detection', frame)
        if cv2.waitKey(5) & 0xFF == ord('q'):
            break

    os.system('pause'

以上是关于openvino +yolov5 自己训练模型并测试的主要内容,如果未能解决你的问题,请参考以下文章

图像识别之Yolov5训练自己的模型

[课程][原创]使用yolov5训练自己实例分割模型windows版

深度学习目标检测---使用yolov5训练自己的数据集模型(Windows系统)

笔记2:yolov5训练自己的目标检测模型_创建并划分数据集

全流程从头在树莓派4B上部署自己训练的yolov5模型(配合NCS2加速)

手把手教你使用YOLOV5训练自己的目标检测模型-口罩检测-视频教程