def verification(model):
dict_content = {'FileName': 'Code'}
dataset = ImageFolder('C:\CR_\data\Test', split=1, mod='verification')
trainloader = torch.utils.data.DataLoader(dataset,
batch_size=1,
shuffle=False,
num_workers=0,
pin_memory=True)
for i, data in enumerate(trainloader, 0):
# 取出一张图片
inputs, file_name = data
"""
1. 将图片进行判断是否含有其他的物体
2. 使用阈值将图片进行除云以外的物体过滤，获得只含云的图片
3. 将图片按照300*300的窗口进行裁剪，将不含云的图片区域按照策略进行舍去
4. 将切割后的图片进行预测
5. 对多个预测结果进行累加取最大值作为未切割图片的预测值
"""
dict_content[file_name[0]] = cutPictureAndComputResult(inputs, model)
print(dict_content)
"""
写入文件
"""
csvFile3 = open('C:\\CR_\\Train_label.csv', 'w', newline='')
writer2 = csv.writer(csvFile3)
for key in dict_content.keys():
writer2.writerow([key, dict_content[key]])
csvFile3.close()
"""
def test_my(path):
x = ImageFolder(path)
imgs = []
for i in x:
imgs.append(np.array(i))
mean, std = get_inception_score(imgs)
print("score = {} +- {}".format(mean, std))
count = count + 1
f.close()
MAE = all_mae / count
Bias = bias / count
RMSE = all_mse / count
RMSE = np.sqrt(RMSE)
# print(MAE, RMSE, Bias)
return MAE, RMSE, Bias
if __name__ == '__main__':
imgsize = 170
path_ = os.path.abspath(r'D:\1WXJ\DATA\CLASS_Japan\devide_3\train')
# trainset = ImageFolder(path_+'/1and2_256/', imgsize, transform)#1and2_256_256_ZJ
trainset = ImageFolder(path_ + '/train_Z_256_35281/', imgsize, transform)
trainloader = torch.utils.data.DataLoader(trainset,
batch_size=128,
shuffle=True,
num_workers=1)
testset = ImageFolder(
r'D:\1WXJ\DATA\CLASS_Japan\devide_3\test\2017-2019\Z', imgsize,
transform)
testloader = torch.utils.data.DataLoader(testset,
batch_size=1,
num_workers=1)
pathoutput = r"D:\1WXJ\Estimate\Model_2\Z_Net_170" #
pathlosssave = os.path.join(r'D:\1WXJ\Estimate\plot_2\Z_Net_170') #
tys_time = {} # map typhoon-time to wind
totalloss = []
test_allloss = []
f.close()
MAE = all_mae / count
Bias = bias / count
RMSE = all_mse / count
RMSE = np.sqrt(RMSE)
# print(MAE, RMSE, Bias)
return MAE, RMSE, Bias
if __name__ == '__main__':
imgsize = 256
path_ = os.path.abspath(r'D:\1WXJ\DATA\WXJ_images\train_leibie')
# trainset = ImageFolder(path_+'/1and2_256/', imgsize, transform)#1and2_256_256_ZJ
trainset = ImageFolder(path_ + '/4and5_256_ZJ_really/', imgsize, transform)
trainloader = torch.utils.data.DataLoader(trainset,batch_size=16, shuffle=True, num_workers=1)
testset = ImageFolder(r'D:\1WXJ\DATA\CLASS_Japan\devide_3\test\2017-2019\Delete_5_zong212_typhoon\A_2',imgsize, transform)
testloader = torch.utils.data.DataLoader(testset, batch_size=1, num_workers=1)
pathoutput = r"D:\1WXJ\Estimate\Model_2\Res_A_170_2_ZJ_0.0005"#
pathlosssave = os.path.join(r'D:\1WXJ\Estimate\plot_2\Res_A_170_2_ZJ_0.0005')#
tys_time = {} # map typhoon-time to wind
totalloss = []
test_allloss = []
max_RMSE = 0
if not os.path.exists(pathlosssave):
os.makedirs(pathlosssave)
if not os.path.exists(pathoutput):
os.makedirs(pathoutput)
model_path = r"D:\1WXJ\Estimate\Model\MODEL_49946_Res34/"
net = resnet34(pretrained=False, modelpath=model_path, num_classes=1000) # batch_size=120, 1GPU Memory < 7000M
import os
from my_transform import demension_reduce
from my_image_folder import ImageFolder
from torch.autograd import Variable
from my_transform import transform
from define_net import Net
from torch.autograd import Variable
if __name__ == '__main__':
path_ = os.path.abspath('.')
net = Net()
net.load_state_dict(torch.load(path_+'/net_relu.pth')) # your net
testset = ImageFolder(path_+'/test_set/',transform) # your test set
f = open(path_+'/result_relu.txt','w') # where to write answer
tys = {} # map typhoon to its max wind
tys_time = {} # map typhoon-time to wind
for i in range(0,testset.__len__()):
image, actual = testset.__getitem__(i)
image = image.expand(1,image.size(0),image.size(1),image.size(2)) # a batch with 1 sample
name = testset.__getitemName__(i)
output = net(Variable(image))
wind = output.data[0][0] # output is a 1*1 tensor
import torch.utils.data
# def testset_loss(testloader, network):
# # loader = torch.utils.data.DataLoader(dataset,batch_size=8,num_workers=2)
# net.eval()
# all_loss = 0.0
# for i,data in enumerate(testloader,0):
# inputs,labels = data
# inputs, labels = Variable(inputs.cuda()), Variable(labels.cuda())
# outputs = network(inputs)
# all_loss = all_loss + abs(labels[0]-outputs.data[0][0])
# return all_loss/i
# net.train()
if __name__ == '__main__':
#path_ = os.path.abspath('.')
model_path = r"D:\1WXJ\Estimate\Model\MODEL_49946_Res34/"
trainset = ImageFolder(r'D:\五分類train\4and5/', transform)
trainloader = torch.utils.data.DataLoader(trainset,
batch_size=8,
shuffle=True,
num_workers=2)
testset = ImageFolder(
r'D:\1WXJ\DATA\CLASS_Japan\devide_3\test\predict_devide\3and4',
transform)
testloader = torch.utils.data.DataLoader(testset,
batch_size=1,
num_workers=2)
pathoutput = r"D:\1WXJ\Estimate\Model\MODEL_45_Res34"
pathlosssave = os.path.join(r'D:\1WXJ\Estimate\plot\plot_45_Res34')
totalloss = []
test_allloss = []
tys_time = {} # map typhoon-time to wind
for i, data in enumerate(loader, 0):
inputs, labels = data
inputs = Variable(inputs)
labels = labels.squeeze()
outputs = net(inputs.cuda())
all_loss = all_loss + function_loss_L1(outputs.cpu(),
labels.long())
return all_loss / i
if __name__ == "__main__":
start_time = time.clock()
losses_his = [[], []]
train_acc = 0
# 数据集加载
trainset = ImageFolder('C:\CR_\data\img', split=0.8, mod='train')
trainloader = torch.utils.data.DataLoader(trainset,
batch_size=100,
shuffle=True,
num_workers=5,
pin_memory=True)
testset = ImageFolder('C:\CR_\data\img', split=0.8, mod='test')
# 模型加载
net = Net()
#net = models.densenet121() #pretrained=True
# 使用CUDA训练模型
if torch.cuda.is_available():
net.cuda()
#初始化权重
nn.init.xavier_uniform_(net.conv1[0].weight.data, gain=1)
nn.init.constant_(net.conv1[0].bias.data, 0.1)
all_loss = 0.0
for i,data in enumerate(loader,0):
inputs,labels = data
inputs = Variable(inputs)
outputs = network(inputs)
all_loss = all_loss + abs(labels[0]-outputs.data[0][0])
return all_loss/i
if __name__ == '__main__':
path_ = os.path.abspath('.')
trainset = ImageFolder(path_+'/train_set/',transform)
trainloader = torch.utils.data.DataLoader(trainset,batch_size=8,
shuffle=True,num_workers=2)
testset = ImageFolder(path_+'/test_set/',transform)
net = Net()
init.xavier_uniform(net.conv1.weight.data,gain=1)
init.constant(net.conv1.bias.data,0.1)
init.xavier_uniform(net.conv2.weight.data,gain=1)
init.constant(net.conv2.bias.data,0.1)
#net.load_state_dict(torch.load(path_+'net_relu.pth'))
print net
criterion = nn.L1Loss()
optimizer = optim.Adam(net.parameters(),lr=0.001)
from define_net_WXJ import Net
from define_net_WXJ import Net_20
from define_net_WXJ import Net_80
from torch.autograd import Variable
if __name__ == '__main__':
path_ = os.path.abspath('.')
net = Net_80()
net.load_state_dict(
torch.load(
r'D:\1WXJ\Estimate\Model\WXJNet_3_256_80/70_net.pth')) # your net
testset = ImageFolder(
r'D:\1WXJ\DATA\CLASS_Japan\devide_3\test\predict_devide\3and4_256_80/',
transform) # your test set
f = open(r'D:\1WXJ\Estimate\plot\WXJNet_3_256_80_JH2/result_3_70.txt'
) # where to write answer
tys = {} # map typhoon to its max wind
tys_time = {} # map typhoon-time to wind
for i in range(0, testset.__len__()):
image, actual = testset.__getitem__(i)
image = image.expand(1, image.size(0), image.size(1),
image.size(2)) # a batch with 1 sample
name = testset.__getitemName__(i)
