def main():
train_list = make_datapath_list(phase="train")
val_list = make_datapath_list(phase="val")
# Dataset
train_dataset = MyDataset(train_list,
transform=ImageTransform(resize, mean, std),
phase="train")
val_dataset = MyDataset(val_list,
transform=ImageTransform(resize, mean, std),
phase="val")
train_dataloader = torch.utils.data.DataLoader(train_dataset,
batch_size=4,
shuffle=True)
val_dataloader = torch.utils.data.DataLoader(val_dataset,
batch_size=4,
shuffle=False)
dataloader_dict = {"train": train_dataloader, "val": val_dataloader}
# Network
use_pretrained = "true"
net = models.vgg16(pretrained=use_pretrained)
net.classifier[6] = nn.Linear(in_features=4096, out_features=2)
# Loss
criterior = nn.CrossEntropyLoss()
# Optimizer
# param_to_update = []
# update_param_name = ["classifier.6.weight", "classifier.6.bias"]
# for name, param in net.named_parameters():
# if name in update_param_name:
# param.requires_grad = True
# param_to_update.append(param)
# print(name)
# else:
# param.requires_grad = False
params1, params2, params3 = param_to_update(net)
optimizer = optim.SGD([
{
'params': params1,
'lr': 1e-4
},
{
'params': params2,
'lr': 5e-4
},
{
'params': params3,
'lr': 1e-3
},
],
momentum=0.9)
train_model(net, dataloader_dict, criterior, optimizer, num_epochs)
def main() :
train_list = make_datapath_list("train")
val_list = make_datapath_list("val")
#dataset
train_dataset = MyDataset(train_list,transform=ImageTransform(resize,mean,std),phase='train')
val_dataset = MyDataset(val_list,transform=ImageTransform(resize,mean,std),phase='val')
#dataloader
train_dataloader = torch.utils.data.DataLoader(train_dataset,batch_size,shuffle=True)
val_dataloader = torch.utils.data.DataLoader(val_dataset,batch_size,shuffle=False)
dataloader_dict = {"train":train_dataloader,"val":val_dataloader}
#NETWORK
use_pretrained = True
net = models.vgg16(pretrained = use_pretrained)
net.classifier[6] = nn.Linear(in_features=4096, out_features=2,bias=True)
print(net)
#setting mode
net = net.train()
#LOSS
criterior = nn.CrossEntropyLoss()
#OPTIMIZER
# Update thong so mong muon
# params_to_update = []
# update_params_name = ["classifier.6.weight","classifier.6.bias"]
# for name,param in net.named_parameters() :
# if name in update_params_name :
# param.requires_grad = True
# params_to_update.append(param)
# print(name)
# else:
# param.requires_grad = False
params1,params2,params3 = params_to_update(net)
# params = update trong so luu vao
#lr = he so hoc
#momentun =
optimizer = optim.SGD([
{"params" :params1, "lr" : 1e-4},
{"params" :params2, "lr" : 5e-4},
{"params" :params3, "lr" : 1e-3}
],momentum = 0.9)
#training
train_model(net,dataloader_dict,criterior, optimizer, num_epoch)
def main():
train_list = make_datapath_list("train")
val_list = make_datapath_list("val")
# dataset
train_dataset = MyDataset(train_list,
transform=ImageTransform(resize, mean, std),
phase="train")
val_dataset = MyDataset(val_list,
transform=ImageTransform(resize, mean, std),
phase="val")
# dataloader
train_dataloader = torch.utils.data.DataLoader(train_dataset,
batch_size,
shuffle=True)
val_dataloader = torch.utils.data.DataLoader(val_dataset,
batch_size,
shuffle=False)
dataloader_dict = {"train": train_dataloader, "val": val_dataloader}
# network
use_pretrained = True
net = models.vgg16(pretrained=use_pretrained)
net.classifier[6] = nn.Linear(in_features=4096, out_features=2)
# loss
criterior = nn.CrossEntropyLoss()
# optimizer
params1, params2, params3 = params_to_update(net)
optimizer = optim.SGD([
{
'params': params1,
'lr': 1e-4
},
{
'params': params2,
'lr': 5e-4
},
{
'params': params3,
'lr': 1e-3
},
],
momentum=0.9)
# training
train_model(net, dataloader_dict, criterior, optimizer, num_epochs)
def make_file_list(self):
file_list = []
path = self.params["data_path"][self.phase]
tissue_path = self.params["data_path"]["tissue_"+self.phase]
if path:
file_list.extend(make_datapath_list(path, self.labels))
if tissue_path:
file_list.extend(make_datapath_list(tissue_path, self.labels))
arrange = self.params["imbalance"]
# データ数の調整ありの場合
if ((arrange=="oversampling") or (arrange=="undersampling")) and (self.phase == "train"):
arrange_file_list = []
file_dict = self.make_file_dict(file_list)
# undrersampling(+bagging)を行う場合
if arrange == "undersampling":
min_file_num = float("inf")
for val in file_dict.values():
min_file_num = min(min_file_num, len(val))
for val in file_dict.values():
# データの重複あり(baggingする場合はこっち)
# arrange_file_list.append(random.choices(val, k=min_file_num))
# データの重複なし(baggingしない場合はこっち)
arrange_file_list.append(random.sample(val, min_file_num))
# oversamplingを行う場合
elif arrange == "oversampling":
max_file_num = 0
for val in file_dict.values():
max_file_num = max(max_file_num, len(val))
for val in file_dict.values():
arrange_file_list.append(random.choices(val, k=max_file_num)) # 重複あり
# random.sampleは再標本化後の数値kがもとの要素数より大きいと使えない
file_list = list(itertools.chain.from_iterable(arrange_file_list))
return file_list
SAVEWHITEPATH = SAVEBASEPATH / 'white'
else:
print('Error:正しいパスを入力してください')
sys.exit()
LABELS = ['Normal', 'PTC', 'fvptc', 'ftc', 'med', 'poor', 'und']
SIZE = 224 # リサイズする大きさ
if '03_迅速標本frozen' in str(BASEPATH):
back_ground_color = (221, 207, 220)
threshold = 203
else:
back_ground_color = (234, 228, 224) # 背景とほぼ同じ色
threshold = 220
path_list = make_datapath_list(str(BASEPATH), LABELS)
print('前処理：縮小、正方形化、ほぼ背景のみの画像の削除を行います')
for path in tqdm(path_list):
image = Image.open(path)
result_image = resize(image=image, size=SIZE)
result_image = padding_square(image=result_image,
background_color=back_ground_color)
if is_white_image(result_image, threshold=threshold):
continue
# save_path = pathlib.Path(path.replace(str(BASEPATH), str(SAVEWHITEPATH)))
else:
save_path = pathlib.Path(
path.replace(str(BASEPATH), str(SAVEBASEPATH)))
if not save_path.parent.exists():
save_path.parent.mkdir(parents=True)
result_image.save(save_path)
from LoadData import HymenopteraDataset
from train import train_model
from utils import ImageTransform, make_datapath_list
from torch.utils.data import DataLoader
from torchvision.models import vgg16
import torch.nn as nn
import torch.optim as optim
import torch
resize = 224
mean = (0.485, 0.456, 0.406)
std = (0.229, 0.224, 0.225)
train_list = make_datapath_list(mode='train')
val_list = make_datapath_list(mode='val')
train_dataset = HymenopteraDataset(file_list=train_list,
transform=ImageTransform(resize, mean, std),
mode='train')
val_dataset = HymenopteraDataset(file_list=val_list,
transform=ImageTransform(resize, mean, std),
mode='val')
batch_size = 32
train_dataloader = DataLoader(train_dataset,
batch_size=batch_size,
shuffle=True)
val_dataloader = DataLoader(val_dataset, batch_size=batch_size, shuffle=False)