def load_image(imgfile):
height, width = 700, 700
image = caffe.io.load_image(imgfile)
transformer = caffe.io.Transformer({'data': (1, 3, height, width)})
transformer.set_transpose('data', (2, 0, 1))
transformer.set_mean('data', np.array([104., 117., 123.]))
transformer.set_raw_scale('data', 7.2801098892805181)
transformer.set_channel_swap('data', (2, 1, 0))
image = transformer.preprocess('data', image)
image = image.reshape(1, 3, height, width)
return image
net = caffenet.CaffeNet(protofile)
print(net)
net.load_weights(weightfile)
net.eval()
image = torch.from_numpy(image)
image = Variable(image)
blobs = net(image)
return blobs, net.models
transformer.set_raw_scale('data', 7.2801098892805181)
transformer.set_channel_swap('data', (2, 1, 0))
image = transformer.preprocess('data', image)
image = image.reshape(1, 3, height, width)
return image
net = caffenet.CaffeNet(protofile)
print(net)
net.load_weights(weightfile)
net.eval()
image = torch.from_numpy(image)
image = Variable(image)
blobs = net(image)
return blobs, net.models
protofile = '/models/testpy_val_91_500_pkg.prototxt'
weightfile = '/models/test2.caffemodel'
imgfn = '08127_image.png'
image = torch.from_numpy(load_image(imgfn))
image = Variable(image)
net = caffenet.CaffeNet(protofile)
net.load_weights(weightfile)
# net = caffenet.CaffeNet(protofile)
# print(net)
# net.load_weights(weightfile)
# net.eval()
blobs = net(image)
def do(fname):
net = caffenet.CaffeNet(os.path.join(srcdir, fname))
net.load_weights(
os.path.join(srcdir, 'vgg16_fast_rcnn_iter_80000.caffemodel'))
torch.save(net.state_dict(), '%s.statedict.pth' % fname)
torch.save(net, '%s.full.pth' % fname)
model = alexbn.AlexBN()
elif args.arch == 'googlenet':
import googlenet
model = googlenet.GoogLeNet()
elif args.arch == 'googlenetbn':
import googlenetbn
model = googlenetbn.GoogLeNetBN()
elif args.arch == 'vgg11':
import vgg
model = vgg.VGG11()
elif args.arch == 'vgg16':
import vgg
model = vgg.VGG16()
elif args.arch == 'caffenet':
import caffenet
model = caffenet.CaffeNet()
else:
raise ValueError('Invalid architecture name')
if args.gpu >= 0:
cuda.get_device(args.gpu).use()
model.to_gpu()
# Setup optimizer
optimizer_class = optimizers.MomentumSGD
if args.compress_gradient:
from optimizer_sgd_compress import MomentumSGDCompress
optimizer_class = MomentumSGDCompress
optimizer = optimizer_class(lr=args.lr, momentum=0.9)
optimizer.setup(model)
optimizer.add_hook(chainer.optimizer.WeightDecay(5e-4))
transforms.ToTensor(), # mmb
])
# torch.utils.data.DataLoader
set = myImageFloder(root="./data/PETA dataset",
label="testdata.txt",
transform=mytransform)
imgLoader = torch.utils.data.DataLoader(set,
batch_size=1,
shuffle=True,
num_workers=2)
print len(set)
path = "./checkpoint1/checkpoint_epoch_60"
net = caffenet.CaffeNet()
net.load_state_dict(torch.load(path))
net.eval()
net.cuda()
'''weight = torch.ones(1,26)
weight[0][2] = 0.3 #18-60
weight[0][22] = 0.4 #trousers
criterion = nn.BCEWithLogitsLoss(weight = weight) #TODO:1.learn 2. weight'''
dataiter = iter(imgLoader)
count = 0
TP = [0.0] * 35
P = [0.0] * 35
TN = [0.0] * 35