介紹
YOLO是基於深度學習端到端的實時目標檢測系統,YOLO將目標區域預測和目標類別預測整合於單個神經網絡模型中,實現在準確率較高的情況下快速目標檢測與識別,更加適合現場應用環境。本案例,我們快速實現一個視頻目標檢測功能,實現的具體原理我們將在單獨的文章中詳細介紹。
下載編譯
我們首先下載Darknet開發框架,Darknet開發框架是YOLO大神級作者自己用C語言編寫的開發框架,支持GPU加速,有兩種下載方式:
- 下載Darknet壓縮包
git clone https://github.com/pjreddie/darknet
下載後,完整的文件內容,如下圖所示:
編譯:
cd darknet
# 編譯
make
編譯後的文件內容,如下圖所示:
下載權重文件
我們這裏下載的是“yolov3”版本,大小是200多M,“yolov3-tiny”比較小,30多M。
wget https://pjreddie.com/media/files/yolov3.weights
下載權重文件後,文件內容如下圖所示:
上圖中的“yolov3-tiny.weights”,"yolov2-tiny.weights"是我單獨另下載的。
C語言預測
./darknet detect cfg/yolov3.cfg yolov3.weights data/dog.jpg
如圖所示,我們已經預測出三種類別以及對應的概率值。模型輸出的照片位於darknet根目錄,名字是“predictions.jpg”,如下圖所示:
讓我們打開模型輸出照片看下:
Python語言預測
我們首先需要將“darknet”文件夾內的“libdarknet.so”文件移動到“darknet/python”內,完成後如下圖所示:
我們將使用Darknet內置的“darknet.py”,進行預測。預測之前,我們需要對文件進行修改:
- 默認py文件基於python2.0,所以對於python3.0及以上需要修改print
- 由於涉及到python和C之間的傳值,所以字符串內容需要轉碼
- 使用絕對路徑
修改完成後,如下圖所示:
打開“darknet/cfg/coco.data”文件,將“names”也改爲絕對路徑(截圖內沒有修改,讀者根據自己的實際路徑修改):
我們可以開始預測了,首先進入“darknet/python”然後執行“darknet.py”文件即可:
結果如下圖所示:
對模型輸出的結果做個簡單的說明,如:
# 分別是:類別,識別概率,識別物體的X座標,識別物體的Y座標,識別物體的長度,識別物體的高度
(b'dog', 0.999338686466217, (224.18377685546875, 378.4237060546875, 178.60214233398438, 328.1665954589844)
視頻檢測
from ctypes import *
import random
import cv2
import numpy as np
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("../python/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 convertBack(x, y, w, h):
xmin = int(round(x - (w / 2)))
xmax = int(round(x + (w / 2)))
ymin = int(round(y - (h / 2)))
ymax = int(round(y + (h / 2)))
return xmin, ymin, xmax, ymax
def array_to_image(arr):
# need to return old values to avoid python freeing memory
arr = arr.transpose(2,0,1)
c, h, w = arr.shape[0:3]
arr = np.ascontiguousarray(arr.flat, dtype=np.float32) / 255.0
data = arr.ctypes.data_as(POINTER(c_float))
im = IMAGE(w,h,c,data)
return im, arr
def detect(net, meta, image, thresh=.5, hier_thresh=.5, nms=.45):
im, image = array_to_image(image)
rgbgr_image(im)
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):
a = dets[j].prob[0:meta.classes]
if any(a):
ai = np.array(a).nonzero()[0]
for i in ai:
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])
if isinstance(image, bytes): free_image(im)
free_detections(dets, num)
return res
if __name__ == "__main__":
cap = cv2.VideoCapture(0)
ret, img = cap.read()
fps = cap.get(cv2.CAP_PROP_FPS)
net = load_net(b"/Users/xiaomingtai/darknet/cfg/yolov2-tiny.cfg", b"/Users/xiaomingtai/darknet/yolov2-tiny.weights", 0)
meta = load_meta(b"/Users/xiaomingtai/darknet/cfg/coco.data")
cv2.namedWindow("img", cv2.WINDOW_NORMAL)
while(True):
ret, img = cap.read()
if ret:
r = detect(net, meta, img)
for i in r:
x, y, w, h = i[2][0], i[2][17], i[2][18], i[2][19]
xmin, ymin, xmax, ymax = convertBack(float(x), float(y), float(w), float(h))
pt1 = (xmin, ymin)
pt2 = (xmax, ymax)
cv2.rectangle(img, pt1, pt2, (0, 255, 0), 2)
cv2.putText(img, i[0].decode() + " [" + str(round(i[1] * 100, 2)) + "]", (pt1[0], pt1[1] + 20), cv2.FONT_HERSHEY_SIMPLEX, 1, [0, 255, 0], 4)
cv2.imshow("img", img)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
模型輸出結果:
模型視頻檢測結果:
沒有GPU的條件下還是不要選擇yolov3了,很慢。
總結
本篇文章主要是YOLO快速上手,我們通過很少的代碼就能實現不錯的目標檢測。當然,想熟練掌握YOLO,理解背後的原理是十分必要的,下篇文章將會重點介紹YOLO原理。