其實現在用的最多的是faster rcnn，等下再弄專案~~~
%matplotlib inlineimport
gluonbook as gb
from mxnet import
autograd,gluon,image,init,nd,contrib
from mxnet.gluon import
loss as gloss,nn
import
time
#類別**層
defcls_predictor(num_anchors,num_classes):
return nn.conv2d(num_anchors*(num_classes+1),kernel_size=3,padding=1)
#邊框**層
defbbox_predictor(num_anchors):
return nn.conv2d(num_anchors*4,kernel_size=3,padding=1)
#鏈結多尺度
defforward(x,block):
block.initialize()
return
block(x)
y1 = forward(nd.zeros((2,8,20,20)),cls_predictor(5,10))
y2 = forward(nd.zeros((2,16,10,10)),cls_predictor(3,10))
y1.shape,y2.shape
defflatten_pred(pred):
return pred.transpose((0,2,3,1)).flatten()
defconcat_preds(preds):
return nd.concat(*[flatten_pred(p) for p in preds],dim=1)
concat_preds([y1,y2]).shape
#減半模組
defdown_sample_blk(num_channels):
blk =nn.sequential()
for _ in range(2):
blk.add(nn.conv2d(num_channels,kernel_size=3,padding=1),
nn.batchnorm(in_channels=num_channels),
nn.activation(
'relu'))
blk.add(nn.maxpool2d(2))
return
blkblk = down_sample_blk(10)
blk.initialize()
x = nd.zeros((2,3,20,20))
y =blk(x)
y.shape
#主體網路塊
defbase_net():
blk =nn.sequential()
for num_filters in [16,32,64]:
blk.add(down_sample_blk(num_filters))
return
blkbnet =base_net()
bnet.initialize()
x = nd.random.uniform(shape=(2,3,256,256))
y =bnet(x)
y.shape
#完整的模型
defget_blk(i):
if i==0: #
0 基礎網路模組
blk =base_net()
elif i==4: #
4 全域性最大池化層模組，將高寬降到1
blk =nn.globalmaxpool2d()
else: #
1 ,2 ,3 高寬減半模組
blk = down_sample_blk(128)
return
blkdef
blk_forward(x,blk,size,ratio,cls_predictor,bbox_predictor):
y =blk(x)
anchors = contrib.nd.multiboxprior(y,sizes=size,ratios=ratio)
cls_preds =cls_predictor(y)
bbox_preds =bbox_predictor(y)
return
(y, anchors, cls_preds,bbox_preds)
sizes = [[0.2, 0.272], [0.37, 0.447], [0.54, 0.619], [0.71, 0.79],
[0.88, 0.961]]
ratios = [[1, 2, 0.5]] * 5num_anchors = len(sizes[0]) + len(ratios[0]) - 1
#完整的tinyssd
class
tinyssd(nn.block):
def__init__(self, num_classes, **kwargs):
super(tinyssd, self).
__init__(**kwargs)
self.num_classes =num_classes
for i in range(5):
#賦值語句 self.blk_i = get_blk(i)
setattr(self, '
blk_%d
' %i,get_blk(i))
setattr(self,
'cls_%d
' %i,cls_predictor(num_anchors,num_classes))
setattr(self,
'bbox_%d
' %i,bbox_predictor(num_anchors))
defforward(self, x):
anchors, cls_preds, bbox_preds = [none]*5,[none]*5,[none]*5
for i in range(5):
#getattr(self, 'blk_%d' % i ) 即訪問 self.blk_i
x, anchors[i], cls_preds[i], bbox_preds[i] =blk_forward(
x, getattr(self,
'blk_%d
' %i), sizes[i], ratios[i],
getattr(self,
'cls_%d
' % i), getattr(self, '
bbox_%d
' %i))
return (nd.concat(*anchors, dim=1),
concat_preds(cls_preds).reshape(
(0, -1, self.num_classes + 1)), concat_preds(bbox_preds))
#測試形狀
net = tinyssd(num_classes=1)
net.initialize()
x = nd.zeros((32,3,256,256))
anchors, cls_preds, bbox_preds =net(x)
print('
output anchors:
',anchors.shape)
print('
output class preds:
',cls_preds.shape)
print('
output bbox preds:
',bbox_preds.shape)
單階多層檢測器 SSD（一）
對於物體檢測任務，第4章的faster rcnn演算法採用了兩階的檢測架構，即首先利用rpn網路進行感興趣區域生成，然後再對該區域進行類別的分類與位置的回歸，這種方法雖然顯著提公升了精度，但也限制了檢測速度。yolo演算法利用回歸的思想，使用一階網路直接完成了物體檢測，速度很快，但是精度有了明顯的下...
單網絡卡多網路
需求 有幾個業務的系統在機房。我自己的電腦經樓層交換機在辦公室。業務系統裡面有獨立的區域網。我想在我辦公室的電腦裡面直接管理這個幾個業務系統。我電腦所在網路與業務系統自身網路相互獨立，且沒有任何物理連線。我也無法將業務系統的網路接入到辦公網路的核心交換機上 我也不想這麼幹 有乙個有利的優勢是，機房跟...
單頁面多頁面區別
單頁面與多頁面的區別 單頁面多頁面 組成 有乙個主要頁面和其他片段頁面組成 由多個完整頁面組成 css js公用載入 每個頁面都分開載入 url模式 a.com pageone a.com pagetwo url模式 a.com pageone.html a.com pagetwo.html 重新整...