ROS Python中深度学习前向传播速度变慢
问题描述：
yolo检测算法，在非ROS环境下前向传播时间0.019s左右，放在ros环境下就1.3s了。慢的难以忍受
分析原因：
因为前向传播在订阅的回调函数中实现，出现了进程问题，所以变慢了。需要重开进程
解决：
订阅函数保存全局图像，新开进程进行前向传播
关键语句：
（1）
import thread
def call_rosspin():
print ('running rosspin...')
rospy.spin()
thread.start_new_thread(call_rosspin, ())
（2）订阅函数写为保存全局图像
def img_L_sub(imgmsg):
global img_L
global flag_img_L
bridge = CvBridge()
img_L = bridge.imgmsg_to_cv2(imgmsg, "bgr8")
flag_img_L = True
print 'img_L recieved'
（3）实例程序
#!/usr/bin/env python
import numpy as np
import cv2
import os, sys
import rospy
from sensor_msgs.msg import Image
from cv_bridge import CvBridge
import thread
from datetime import datetime
from yolo_ros.msg import TargetsInfo
#sys.path.insert(0, '../../python/')
caffe_root = '/home/alvin/Desktop/caffe-yolo/'
sys.path.insert(0, caffe_root + 'python')
import caffe
caffe.set_device(0)
caffe.set_mode_gpu()
mean = np.require([104, 117, 123], dtype=np.float32)[:, np.newaxis, np.newaxis] ##??[104, 117, 123] [ r g b]???
img_L = cv2.imread('Null')
flag_img_L = False
pb = 0.3
def img_L_sub(imgmsg):
global img_L
global flag_img_L
bridge = CvBridge()
img_L = bridge.imgmsg_to_cv2(imgmsg, "bgr8")
flag_img_L = True
print 'img_L recieved'
def call_rosspin():
print ('running rosspin...')
rospy.spin()
def nms(dets, thresh):
# -------------------------
# Pure Python NMS baseline.
# Written by Ross Girshick
# -------------------------
x1 = dets[:, 0] - dets[:, 2] / 2.
y1 = dets[:, 1] - dets[:, 3] / 2.
x2 = dets[:, 0] + dets[:, 2] / 2.
y2 = dets[:, 1] + dets[:, 3] / 2.
scores = dets[:, 4]
areas = (x2 - x1 + 1) * (y2 - y1 + 1)
order = scores.argsort()[::-1]
keep = []
while order.size > 0:
i = order[0]
keep.append(i)
xx1 = np.maximum(x1[i], x1[order[1:]])
yy1 = np.maximum(y1[i], y1[order[1:]])
xx2 = np.minimum(x2[i], x2[order[1:]])
yy2 = np.minimum(y2[i], y2[order[1:]])
w = np.maximum(0.0, xx2 - xx1 + 1)
h = np.maximum(0.0, yy2 - yy1 + 1)
inter = w * h
ovr = inter / (areas[i] + areas[order[1:]] - inter)
inds = np.where(ovr <= thresh)[0]
order = order[inds + 1]
return dets[np.require(keep), :]
def parse_result(reg_out):
num_classes = 20
num_objects = 2
side = 7
locations = side ** 2
boxes = np.zeros((num_objects * locations, 6), dtype=np.float32)
for i in range(locations):
tmp_scores = reg_out[i:num_classes*locations:locations]
max_class_ind = np.argsort(tmp_scores)[-1]
max_prob = np.max(tmp_scores)
obj_index = num_classes * locations + i
obj_scores = max_prob * reg_out[obj_index:(obj_index+num_objects*locations):locations]
coor_index = (num_classes + num_objects) * locations + i
for j in range(num_objects):
boxes[i*num_objects+j][5] = max_class_ind
boxes[i*num_objects+j][4] = obj_scores[j]
box_index = coor_index + j * 4 * locations
boxes[i*num_objects+j][0] = (i % side + reg_out[box_index + 0 * locations]) / float(side)
boxes[i*num_objects+j][1] = (i / side + reg_out[box_index + 1 * locations]) / float(side)
boxes[i*num_objects+j][2] = reg_out[box_index + 2 * locations] ** 2
boxes[i*num_objects+j][3] = reg_out[box_index + 3 * locations] ** 2
return nms(boxes, 0.7)
def show_boxes(im, boxes, thresh=0.7, show=0):
print boxes.shape
#im = cv2.imread(im_path)
ori_w = im.shape[1]
ori_h = im.shape[0]
for box in boxes:
if box[4] < thresh: #box[4] is the possibality of some class
continue
print box
print box[5]
box = box[:4] # [:4]means what??
x1 = max(0, int((box[0] - box[2] / 2.) * ori_w))
y1 = max(0, int((box[1] - box[3] / 2.) * ori_h))
x2 = min(ori_w - 1, int((box[0] + box[2] / 2.) * ori_w))
y2 = min(ori_h - 1, int((box[1] + box[3] / 2.) * ori_h))
cv2.rectangle(im, (x1, y1), (x2, y2), (0, 0, 255), 2)
return im
def det(model, im_path, show=0):
'''forward processing'''
im = cv2.imread(im_path)
im = cv2.resize(im, (448, 448))
im = np.require(im.transpose((2, 0, 1)), dtype=np.float32)##?? rearrgage r g b channel??
im -= mean # substract mean
model.blobs['data'].data[...] = im # ... means what??
out_blobs = model.forward() #out 1470
reg_out = out_blobs["regression"] #regression what kind of keyword??
boxes = parse_result(reg_out[0])
show_boxes(im_path, boxes, 0.2)
if __name__=="__main__":
rospy.init_node('yolo_uav', anonymous=True)
sub_img_L = rospy.Subscriber("/usb_cam/image_raw", Image, img_L_sub)
targets_Pub = rospy.Publisher("/yolo/targets", TargetsInfo, queue_size=80)
thread.start_new_thread(call_rosspin, ())
net_proto = "./net/gnet_deploy.prototxt"
model_path = "./models/gnet_yolo_iter_12000.caffemodel"
model = caffe.Net(net_proto, model_path, caffe.TEST)
TargetsInfo_t = TargetsInfo()
key = 100
thresh = 0.2
while not rospy.is_shutdown():
if flag_img_L == 0:
continue
flag_img_L = 0
start = datetime.now()
TargetsInfo_t.targetNum = 0
img = img_L
im = cv2.resize(img,(448,448))
im = np.require(im.transpose((2, 0, 1)), dtype=np.float32)
im -= mean
model.blobs['data'].data[...] = im # ... means what??
out_blobs = model.forward()
reg_out = out_blobs['regression']
boxes = parse_result(reg_out[0])
im = img
print boxes.shape
ori_w = im.shape[1]
ori_h = im.shape[0]
for box in boxes:
if box[4] < pb: #box[4] is the possibality of some class
continue
i = int(TargetsInfo_t.targetNum)
#box[5] #[x2 x1 y2 y1 thresh ID]
x1 = max(0, int((box[0] - box[2] / 2.) * ori_w))
y1 = max(0, int((box[1] - box[3] / 2.) * ori_h))
x2 = min(ori_w - 1, int((box[0] + box[2] / 2.) * ori_w))
y2 = min(ori_h - 1, int((box[1] + box[3] / 2.) * ori_h))
cv2.rectangle(im, (x1, y1), (x2, y2), (0, 0, 255), 2)
TargetsInfo_t.targetId[i] = int(box[5])
TargetsInfo_t.probability[i] = float(box[4])
TargetsInfo_t.x1[i] = x1
TargetsInfo_t.y1[i] = y1
TargetsInfo_t.x2[i] = x2
TargetsInfo_t.y2[i] = y2
TargetsInfo_t.targetNum = TargetsInfo_t.targetNum +1
targets_Pub.publish(TargetsInfo_t)
end = datetime.now()
elapsedTime = end-start
elapsedTime = elapsedTime.total_seconds()*1000
elapsedTime = 1000 / elapsedTime
print 'TargetsInfo_t.targetNum',TargetsInfo_t.targetNum
print 'FPS: ', elapsedTime
cv2.imshow('out_2', im)
key = cv2.waitKey(3)