HED算法介紹

代碼演示

自定義層解析的代碼如下：

#自定義層
classCropLayer(object):
def__init__(self,params,blobs):
self.xstart=0
self.xend=0
self.ystart=0
self.yend=0

defgetMemoryShapes(self,inputs):
inputShape,targetShape=inputs[0],inputs[1]
batchSize,numChannels=inputShape[0],inputShape[1]
height,width=targetShape[2],targetShape[3]

self.ystart=(inputShape[2]-targetShape[2])//2
self.xstart=(inputShape[3]-targetShape[3])//2
self.yend=self.ystart+height
self.xend=self.xstart+width
return[[batchSize,numChannels,height,width]]

defforward(self,inputs):
return[inputs[0][:,:,self.ystart:self.yend,self.xstart:self.xend]]

加載網(wǎng)絡(luò)，進(jìn)行多全層次的邊緣檢測(cè)：

#自定義層
#注冊(cè)自定義層
cv.dnn_registerLayer('Crop',CropLayer)

#Loadthemodel.
net=cv.dnn.readNet("D:/projects/models/hed/deploy.prototxt","D:/projects/models/hed/hed_pretrained_bsds.caffemodel")

kWinName='Holistically-NestedEdgeDetection'
cv.namedWindow('Input',cv.WINDOW_AUTOSIZE)
cv.namedWindow(kWinName,cv.WINDOW_AUTOSIZE)

cap=cv.VideoCapture(0)
whilecv.waitKey(1)0:
hasFrame,frame=cap.read()
frame=cv.flip(frame,1)
ifnothasFrame:
cv.waitKey()
break
cv.imshow('Input',frame)

inp=cv.dnn.blobFromImage(frame,scalefactor=1.0,size=(500,500),
mean=(104.00698793,116.66876762,122.67891434),
swapRB=False,crop=False)
net.setInput(inp)
out=net.forward()
out=out[0,0]
out=cv.resize(out,(frame.shape[1],frame.shape[0]))
cv.imshow(kWinName,out)

運(yùn)行效果

圖像測(cè)試

視頻測(cè)試

論文與源碼

源碼與預(yù)訓(xùn)練模型
https://github.com/s9xie/hed
論文地址：
https://arxiv.org/abs/1504.06375

審核編輯 ：李倩