机器人

ROS机器人Diego 1#制作（三十四）机器视觉 - 利用人工智能风格迁移技术拍摄不同画风的视频

robot
2020年4月30日
599

风格迁移，就是将一种图片的风格迁移到其他图片上，改变其他图片的风格，很好玩的一个人工自能模型，github上已经有很多实现的方法，本文参考https://github.com/hzy46/fast-neural-style-tensorflow 的算法，利用Diego1#的平台实现实时视频的风格转换，先上两张图看效果：
![这里写图片描述](https://imgconvert.csdnimg.cn/aHR0cDovL2ltZy5ibG9nLmNzZG4ubmV0LzIwMTcwNzI4MjAyNDMyMjQz?x-oss-process=image/format,png)
![这里写图片描述](https://imgconvert.csdnimg.cn/aHR0cDovL2ltZy5ibG9nLmNzZG4ubmV0LzIwMTcwNzI4MjAyNDUyNTA2?x-oss-process=image/format,png)
是不是很酷呢，其实实现方法和上篇博文中的原理是一样的，只是把人工智能的算法包装成一个ROS Node，把结果发布为Compressed Image，所以这里只把node节点的代码贴出来，原理就不在讲解了，模型的源代码可以到上面提到的github上去下载

```
这里写代码片#!/usr/bin/env python

import rospy
from sensor_msgs.msg import Image as ROSImage
from sensor_msgs.msg import CompressedImage as ROSImage_C
from cv_bridge import CvBridge
import numpy as np
import os
import tensorflow as tf
import PIL.Image as Image
import time
import StringIO

from fast_neural_style.preprocessing import preprocessing_factory
from fast_neural_style import reader
from fast_neural_style import model

class fastNeuralStyle():
    def __init__(self):
	rospy.init_node('fastNeuralStyle')
	
	# Set the shutdown function (stop the robot)
        rospy.on_shutdown(self.shutdown)
                
        image_topic = rospy.get_param("~image_topic", "")        
        self.model_file=rospy.get_param("~model_file", "")
        self.loss_model=rospy.get_param("~loss_model", "")
        
        self._cv_bridge = CvBridge()        
        self.vfc=0
	
	self._sub = rospy.Subscriber(image_topic, ROSImage, self.callback, queue_size=1)	
	self._pub = rospy.Publisher('fast_Neural_Style', ROSImage_C, queue_size=1)
	
	rospy.loginfo("initialization has finished...")
	
        
    def callback(self,image_msg):
	if self.vfc<12:
		self.vfc=self.vfc+1
	else:
		self.callbackfun(image_msg)
		self.vfc=0	
		
    def callbackfun(self, image_msg):
	with tf.Graph().as_default():
        	with tf.Session().as_default() as sess:
			 cv_image = self._cv_bridge.imgmsg_to_cv2(image_msg, "bgr8")

pil_img = Image.fromarray(cv_image)			 
			 (im_width, im_height) = pil_img.size	
			 
			 		 
			 # Read image data.
			 image_preprocessing_fn, _ = preprocessing_factory.get_preprocessing(
			 	self.loss_model,
			 	is_training=False)
			 
			 image_np =np.array(pil_img.getdata()).reshape((im_height, im_width, 3)).astype(np.uint8)
			 
			 image_np=image_preprocessing_fn(image_np,im_height,im_width)

# Add batch dimension
			 image_np = tf.expand_dims(image_np, 0)

generated = model.net(image_np, training=False)
			 generated = tf.cast(generated, tf.uint8)

# Remove batch dimension
			 generated = tf.squeeze(generated, [0])

# Restore model variables.
			 saver = tf.train.Saver(tf.global_variables(), write_version=tf.train.SaverDef.V1)
			 sess.run([tf.global_variables_initializer(), tf.local_variables_initializer()])
			 # Use absolute path
			 self.model_file = os.path.abspath(self.model_file)
			 saver.restore(sess, self.model_file)			 	
			 
			 jpeg_bin = sess.run(tf.image.encode_jpeg(generated))
			 jpeg_str = StringIO.StringIO(jpeg_bin)
			 jpeg_image = Image.open(jpeg_str)
			 
			 img = np.array(jpeg_image, dtype=np.uint8)
			 
			 
			 ros_compressed_image=self._cv_bridge.cv2_to_compressed_imgmsg(img)
			 self._pub.publish(ros_compressed_image)

def shutdown(self):
        rospy.loginfo("Stopping the fastNeuralStyle...")
        rospy.sleep(1) 
        
if __name__ == '__main__':
    try:
        fastNeuralStyle()
        rospy.spin()
    except rospy.ROSInterruptException:
        rospy.loginfo("Ros TensorFlow fastNeuralStyle has started.")
```