前记:
作用说明:学习笔记,主要用于自我记录。
(PS:本人新手,文章仅供参考;如有错误,欢迎各位大神批评指正!)
最近刚刚接触yolo,由于yolo官网和网上各种资料几乎都是基于C语言的,本人觉得python比较简洁,故用python实现了C可实现的部分功能。
该文承接上篇博文“yolo在python下图片检测及画框”。此篇介绍(2)yolo用python实现检测本地视频。
【本系列博文代码基于原作者程序,文中代码由于TAB缩进量可能会出现缩进问题,敬请谅解。】
1、使用环境和平台
ubuntu 14.04+ python2.7+opencv2.4+yolo
2、检测本地视频的python代码
在darknet/python文件夹下新建my_local_video_darknet.py
以下是我实现检测本地视频的代码:
#!coding=utf-8
#modified by Leo at 2018.04.25
#function: 1,detect the video captured by the webcam
# 2,No passing frames, so it's very slow on my computer.
from ctypes import *
import math
import random
#import module named cv2 to draw
import cv2
import Image
def sample(probs):
s = sum(probs)
probs = [a/s for a in probs]
r = random.uniform(0, 1)
for i in range(len(probs)):
r = r - probs[i]
if r <= 0:
return i
return len(probs)-1
def c_array(ctype, values):
arr = (ctype*len(values))()
arr[:] = values
return arr
class BOX(Structure):
_fields_ = [("x", c_float),
("y", c_float),
("w", c_float),
("h", c_float)]
class DETECTION(Structure):
_fields_ = [("bbox", BOX),
("classes", c_int),
("prob", POINTER(c_float)),
("mask", POINTER(c_float)),
("objectness", c_float),
("sort_class", c_int)]
class IMAGE(Structure):
_fields_ = [("w", c_int),
("h", c_int),
("c", c_int),
("data", POINTER(c_float))]
class METADATA(Structure):
_fields_ = [("classes", c_int),
("names", POINTER(c_char_p))]
#lib = CDLL("/home/pjreddie/documents/darknet/libdarknet.so", RTLD_GLOBAL)
lib = CDLL("/home/username/darknet/libdarknet.so", RTLD_GLOBAL)
lib.network_width.argtypes = [c_void_p]
lib.network_width.restype = c_int
lib.network_height.argtypes = [c_void_p]
lib.network_height.restype = c_int
predict = lib.network_predict
predict.argtypes = [c_void_p, POINTER(c_float)]
predict.restype = POINTER(c_float)
set_gpu = lib.cuda_set_device
set_gpu.argtypes = [c_int]
make_image = lib.make_image
make_image.argtypes = [c_int, c_int, c_int]
make_image.restype = IMAGE
get_network_boxes = lib.get_network_boxes
get_network_boxes.argtypes = [c_void_p, c_int, c_int, c_float, c_float, POINTER(c_int), c_int, POINTER(c_int)]
get_network_boxes.restype = POINTER(DETECTION)
make_network_boxes = lib.make_network_boxes
make_network_boxes.argtypes = [c_void_p]
make_network_boxes.restype = POINTER(DETECTION)
free_detections = lib.free_detections
free_detections.argtypes = [POINTER(DETECTION), c_int]
free_ptrs = lib.free_ptrs
free_ptrs.argtypes = [POINTER(c_void_p), c_int]
network_predict = lib.network_predict
network_predict.argtypes = [c_void_p, POINTER(c_float)]
reset_rnn = lib.reset_rnn
reset_rnn.argtypes = [c_void_p]
load_net = lib.load_network
load_net.argtypes = [c_char_p, c_char_p, c_int]
load_net.restype = c_void_p
do_nms_obj = lib.do_nms_obj
do_nms_obj.argtypes = [POINTER(DETECTION), c_int, c_int, c_float]
do_nms_sort = lib.do_nms_sort
do_nms_sort.argtypes = [POINTER(DETECTION), c_int, c_int, c_float]
free_image = lib.free_image
free_image.argtypes = [IMAGE]
letterbox_image = lib.letterbox_image
letterbox_image.argtypes = [IMAGE, c_int, c_int]
letterbox_image.restype = IMAGE
load_meta = lib.get_metadata
lib.get_metadata.argtypes = [c_char_p]
lib.get_metadata.restype = METADATA
load_image = lib.load_image_color
load_image.argtypes = [c_char_p, c_int, c_int]
load_image.restype = IMAGE
rgbgr_image = lib.rgbgr_image
rgbgr_image.argtypes = [IMAGE]
predict_image = lib.network_predict_image
predict_image.argtypes = [c_void_p, IMAGE]
predict_image.restype = POINTER(c_float)
def classify(net, meta, im):
out = predict_image(net, im)
res = []
for i in range(meta.classes):
res.append((meta.names[i], out[i]))
res = sorted(res, key=lambda x: -x[1])
return res
def detect(net, meta, image, thresh=.5, hier_thresh=.5, nms=.45):
im = load_image(image, 0, 0)
num = c_int(0)
pnum = pointer(num)
predict_image(net, im)
dets = get_network_boxes(net, im.w, im.h, thresh, hier_thresh, None, 0, pnum)
num = pnum[0]
if (nms): do_nms_obj(dets, num, meta.classes, nms);
res = []
for j in range(num):
for i in range(meta.classes):
if dets[j].prob[i] > 0:
b = dets[j].bbox
res.append((meta.names[i], dets[j].prob[i], (b.x, b.y, b.w, b.h)))
res = sorted(res, key=lambda x: -x[1])
free_image(im)
free_detections(dets, num)
return res
# 2018.04.25
def showPicResult(image):
img = cv2.imread(image)
cv2.imwrite(out_img, img)
for i in range(len(r)):
x1=r[i][2][0]-r[i][2][2]/2
y1=r[i][2][1]-r[i][2][3]/2
x2=r[i][2][0]+r[i][2][2]/2
y2=r[i][2][1]+r[i][2][3]/2
im = cv2.imread(out_img)
#draw different color rectangle
'''
#random color
r_color = random.randint(0,255)
g = random.randint(0,255)
b = random.randint(0,255)
cv2.rectangle(im,(int(x1),int(y1)),(int(x2),int(y2)),(r_color,g,b),3)
'''
cv2.rectangle(im,(int(x1),int(y1)),(int(x2),int(y2)),(0,255,0),3)
#This is a method that works well.
cv2.imwrite(out_img, im)
cv2.imshow('yolo_image_detector', cv2.imread(out_img))
if __name__ == "__main__":
net = load_net("/home/username/darknet/cfg/yolov3.cfg", "/home/username/darknet/weights/yolov3.weights", 0)
meta = load_meta("/home/username/darknet/cfg/coco.data")
out_img = "/home/username/darknet/data/test_result.jpg"
video_tmp = "/home/username/darknet/data/video_tmp.jpg"
origin_video = '/home/username/darknet/data/video_20180426_test.mp4'
# make a video_object and init the video object
cap = cv2.VideoCapture(origin_video)
# define picture to_down' coefficient of ratio
scaling_factor = 0.5
count = 0
# loop until press 'esc' or 'q'
while (cap.isOpened()):
# collect current frame
ret, frame = cap.read()
if ret == True:
count = count + 1
#print count
else:
break
#detect and show per 50 frames
if count == 50:
count = 0
# resize the frame
frame = cv2.resize(frame,None,fx=scaling_factor,fy=scaling_factor,interpolation=cv2.INTER_AREA)
img_arr = Image.fromarray(frame)
img_goal = img_arr.save(video_tmp)
r = detect(net, meta, video_tmp)
print r
print ''
print '#*********************************#'
#display the rectangle of the objects in window
showPicResult(video_tmp)
else:
continue
# wait 1ms per iteration; press Esc to jump out the loop
c = cv2.waitKey(1)
if (c==27) or (0xFF == ord('q')):
break
# release and close the display_window
cap.release()
如果出现配置问题,可参考第一篇博文。
3、代码解释
除了showPicResult()函数(用作绘制物体框)在上一篇博客中已经加入,此程序中再次加入了读取本地视频的相关函数。
# notice: path of local video
origin_video = '/home/username/darknet/data/video_20180426_test.mp4'
# make a video_object and init the video object
cap = cv2.VideoCapture(origin_video)
# define picture to_down' coefficient of ratio
scaling_factor = 0.5
count = 0
# loop until press 'esc' or 'q'
while (cap.isOpened()):
# collect current frame
ret, frame = cap.read()
if ret == True:
count = count + 1
#print count
else:
break
#detect and show per 50 frames
if count == 50:
count = 0
# resize the frame
frame = cv2.resize(frame,None,fx=scaling_factor,fy=scaling_factor,interpolation=cv2.INTER_AREA)
img_arr = Image.fromarray(frame)
img_goal = img_arr.save(video_tmp)
r = detect(net, meta, video_tmp)
print r
print ''
print '#*********************************#'
#display the rectangle of the objects in window
showPicResult(video_tmp)
else:
continue
# wait 1ms per iteration; press Esc to jump out the loop
c = cv2.waitKey(1)
if (c==27) or (0xFF == ord('q')):
break
# release and close the display_window
cap.release()
此部分主要是读取本地视频的每帧,然后利用python的Image工具把numpy.ndarry的帧数据类型转换为Image类型数据(如不转换在“
im = load_image(image, 0, 0)
”中就会报错);转换为Image数据类型之后,就可利用我们在第一篇博文中提到的showPicResult()函数在视频中显示物体框。
另外:
由于本人电脑没有独显,处理速度慢,所以在程序中跳过了一些帧(每50帧用yolo检测一次);即使有GPU,也没必要每一帧都去检测,每秒几十帧的图像,大多数都是一样的,挑选几帧检测一下即可。
后记:
程序理解起来很简单,在我自己的电脑上完全可以实现,如有问题欢迎各位大佬批评指正!
联系邮箱:2052383522@qq.com