def getDenseNet(num_classes, ctx):
densenet = vision.densenet201(pretrained=True, ctx=ctx)
net = vision.densenet201(classes=num_classes, prefix='densenet0_')
with net.name_scope():
net.output = nn.Dense(num_classes, flatten=True)
net.output.collect_params().initialize(mx.init.Xavier(
rnd_type='gaussian', factor_type="in", magnitude=2),
ctx=ctx)
net.features = densenet.features
net.collect_params().reset_ctx(ctx)
inputs = mx.sym.var('data')
out = net(inputs)
internals = out.get_internals()
outputs = [
internals['densenet0_conv3_fwd_output'],
internals['densenet0_stage4_concat15_output'],
internals['densenet0_dense1_fwd_output']
]
feat_model = gluon.SymbolBlock(outputs,
inputs,
params=net.collect_params())
feat_model._prefix = 'densenet0_'
return feat_model
def load_model():
print("we're loading densenet model: \
https://modelzoo.co/model/densely-connected-convolutional-networks-2")
densenetX = vision.densenet201(pretrained=True)
print("we just loaded: ")
type(densenetX)
return densenetX
def load_model():
print("we're loading densenet model: \
https://modelzoo.co/model/densely-connected-convolutional-networks-2")
densenetX = vision.densenet201(pretrained=True)
print("we just loaded: ")
type(densenetX)
print("Now we're loading YOLO")
yolo_netX = model_zoo.get_model('yolo3_darknet53_voc', pretrained=True)
type(yolo_netX)
return densenetX, yolo_netX
def densenet201mxnetload():
net = vision.densenet201(pretrained=True)
net.hybridize()
return net
from mxnet.gluon import nn
from mxnet.gluon.model_zoo import vision as models
dense_net=models.densenet201(pretrained=False,prefix="a_")
net=nn.HybridSequential(prefix="a_")
for layer in dense_net.features[0:13]:
net.add(layer)
class output_block(nn.HybridBlock):
def __init__(self,**kwargs):
super (output_block,self).__init__(**kwargs)
with self.name_scope():
self.dense1=nn.Dense(256,activation='relu')
self.globalavg=nn.GlobalAvgPool2D()
self.dense2=nn.Dense(1)
def hybrid_forward(self,F,x):
out=self.globalavg(x)
out=self.dense1(out)
out=self.dense2(out)
return out
output=output_block(prefix="a_")
with net.name_scope():
net.add(output)
out = net(data.as_in_context(mx.gpu()))
print(out.shape)
out = out.reshape(
(30720)) # dense169 6656 dense201 7890 //122880
out = out.as_in_context(mx.cpu())
out = out.asnumpy()
res = ''
for item in out:
res += str(item) + ' '
#print(res)
fw1.write(res + '\n')
label = root.strip().split('\\')[-1]
fw2.write(label + '\n')
print(label)
fw1.close()
fw2.close()
print('net done!!!')
pretrained_net = models.densenet201(pretrained=True)
net = nn.HybridSequential()
for layer in pretrained_net.features:
net.add(layer)
#print(net)
ctx = mx.gpu()
net.collect_params().reset_ctx(ctx)
net.hybridize()
dir = '../data/dense1920'
suffix = ['jpg']
batch_net(dir, suffix, net)
