在Python 3中使用YOLOv2

YOLOv2是Joseph Redmon提出的针对YOLO算法不足的改进版本，作者使用了一系列的方法对原来的YOLO多目标检测框架进行了改进，在保持原有速度的优势之下，精度上得以提升，此外作者提出了一种目标分类与检测的联合训练方法，通过这种方法YOLO9000可以同时在COCO和ImageNet数据集中进行训练，训练后的模型可以实现多达9000种物体的实时检测。

Paper：https://arxiv.org/abs/1612.08242
Github：https://github.com/pjreddie/darknet
Website：https://pjreddie.com/darknet/yolo

作者为YOLO算法设计了独有的深度学习框架darknet，因此没有提供Python的接口。在实验中，我找到了两种在Python 3中使用YOLOv2网络的方法。

第一种：为darknet添加Python接口

Github：https://github.com/SidHard/py-yolo2

该项目使用了原始的darknet网络，需要使用cmake重新编译源码，因此在Linux上使用更为方便一些。

首先从git上下载该项目
git clone https://github.com/SidHard/py-yolo2.git
执行cmake生成项目
cmake .. && make
最后执行yolo.py测试项目，相应的网络结构.cfg文件保存在cfg文件夹中，权值.weight文件放在根目录下，这些可以从darknet的官方网站上下载使用。

第二种：使用keras

Github：https://github.com/allanzelener/YAD2K

该项目使用了keras与tensorflow-gpu，因此可以在任何使用该框架的环境下运行，我在自己的程序中使用的该种方法。

首先下载源文件并且配置环境，可以使用anaconda环境或者在全局安装。

git clone https://github.com/allanzelener/yad2k.git

cd yad2k

# [Option 1] To replicate the conda environment:

conda env create -f environment.yml

source activate yad2k

# [Option 2] Install everything globaly.

pip install numpy

pip install tensorflow-gpu  # CPU-only: conda install -c conda-forge tensorflow

pip install keras # Possibly older release: conda install keras

快速开始

• 从Darknet官方下载model：official YOLO website.
  wget http://pjreddie.com/media/files/yolo.weights
• 将 Darknet YOLO_v2 model转换为Keras model.
  ./yad2k.py cfg/yolo.cfg yolo.weights model_data/yolo.h5
• 测试图片位于 images/文件夹.
  ./test_yolo.py model_data/yolo.h5

最后执行test_yolo就可以执行网络，在images/out/文件夹里可以看到执行效果。

dog.jpg

eagle.jpg

giraffe.jpg

horses.jpg

为了方便模型用于测试视频与图片，我对demo做了修改，相比原来的测试代码，能够直接移植到项目中去，对象化的程序也更易于修改，代码如下

#! /usr/bin/env python

"""Run a YOLO_v2 style detection model on test images."""

import cv2

import os

import time

import numpy as np

from keras import backend as K

from keras.models import load_model

from yad2k.models.keras_yolo import yolo_eval, yolo_head

class YOLO(object):

    def __init__(self):

        self.model_path = 'model_data/yolo.h5'

        self.anchors_path = 'model_data/yolo_anchors.txt'

        self.classes_path = 'model_data/coco_classes.txt'

        self.score = 0.3

        self.iou = 0.5

        self.class_names = self._get_class()

        self.anchors = self._get_anchors()

        self.sess = K.get_session()

        self.boxes, self.scores, self.classes = self.generate()

    def _get_class(self):

        classes_path = os.path.expanduser(self.classes_path)

        with open(classes_path) as f:

            class_names = f.readlines()

        class_names = [c.strip() for c in class_names]

        return class_names

    def _get_anchors(self):

        anchors_path = os.path.expanduser(self.anchors_path)

        with open(anchors_path) as f:

            anchors = f.readline()

            anchors = [float(x) for x in anchors.split(',')]

            anchors = np.array(anchors).reshape(-1, 2)

        return anchors

    def generate(self):

        model_path = os.path.expanduser(self.model_path)

        assert model_path.endswith('.h5'), 'Keras model must be a .h5 file.'

        self.yolo_model = load_model(model_path)

        # Verify model, anchors, and classes are compatible

        num_classes = len(self.class_names)

        num_anchors = len(self.anchors)

        # TODO: Assumes dim ordering is channel last

        model_output_channels = self.yolo_model.layers[-1].output_shape[-1]

        assert model_output_channels == num_anchors * (num_classes + 5), \

            'Mismatch between model and given anchor and class sizes'

        print('{} model, anchors, and classes loaded.'.format(model_path))

        # Check if model is fully convolutional, assuming channel last order.

        self.model_image_size = self.yolo_model.layers[0].input_shape[1:3]

        self.is_fixed_size = self.model_image_size != (None, None)

        # Generate output tensor targets for filtered bounding boxes.

        # TODO: Wrap these backend operations with Keras layers.

        yolo_outputs = yolo_head(self.yolo_model.output, self.anchors, len(self.class_names))

        self.input_image_shape = K.placeholder(shape=(2, ))

        boxes, scores, classes = yolo_eval(yolo_outputs, self.input_image_shape, score_threshold=self.score, iou_threshold=self.iou)

        return boxes, scores, classes

    def detect_image(self, image):

        start = time.time()

        y, x, _ = image.shape

        if self.is_fixed_size:  # TODO: When resizing we can use minibatch input.

            resized_image = cv2.resize(image, tuple(reversed(self.model_image_size)), interpolation=cv2.INTER_CUBIC)

            image_data = np.array(resized_image, dtype='float32')

        else:

            image_data = np.array(image, dtype='float32')

        image_data /= 255.

        image_data = np.expand_dims(image_data, 0)  # Add batch dimension.

        out_boxes, out_scores, out_classes = self.sess.run(

            [self.boxes, self.scores, self.classes],

            feed_dict={

                self.yolo_model.input: image_data,

                self.input_image_shape: [image.shape[0], image.shape[1]],

                K.learning_phase(): 0

            })

        print('Found {} boxes for {}'.format(len(out_boxes), 'img'))

        for i, c in reversed(list(enumerate(out_classes))):

            predicted_class = self.class_names[c]

            box = out_boxes[i]

            score = out_scores[i]

            label = '{} {:.2f}'.format(predicted_class, score)

            top, left, bottom, right = box

            top = max(0, np.floor(top + 0.5).astype('int32'))

            left = max(0, np.floor(left + 0.5).astype('int32'))

            bottom = min(y, np.floor(bottom + 0.5).astype('int32'))

            right = min(x, np.floor(right + 0.5).astype('int32'))

            print(label, (left, top), (right, bottom))

            cv2.rectangle(image, (left, top), (right, bottom), (255, 0, 0), 2)

            cv2.putText(image, label, (left, int(top - 4)), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 1, cv2.LINE_AA)

        end = time.time()

        print(end - start)

        return image

    def close_session(self):

        self.sess.close()

def detect_vedio(video, yolo):

    camera = cv2.VideoCapture(video)

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

    while True:

        res, frame = camera.read()

        if not res:

            break

        image = yolo.detect_image(frame)

        cv2.imshow("detection", image)

        if cv2.waitKey(110) & 0xff == 27:

                break

    yolo.close_session()

def detect_img(img, yolo):

    image = cv2.imread(img)

    r_image = yolo.detect_image(image)

    cv2.namedWindow("detection")

    while True:

        cv2.imshow("detection", r_image)

        if cv2.waitKey(110) & 0xff == 27:

                break

    yolo.close_session()

if __name__ == '__main__':

    yolo = YOLO()

    img = 'E:\Documents\Downloads\YAD2K-master\YAD2K-master\images\horses.jpg'

    video = 'E:\Documents\Documents\python\Traffic\data\person.avi'  

    detect_img(img, yolo)

    detect_vedio(video, yolo)

作者：洛荷
链接：https://www.jianshu.com/p/3e77cefeb49b
�碓矗�简书
著作权归作者所有。商业转载请联系作者获得授权，非商业转载请注明出处。