def main():
batch_size = 20
# model = models.resnet18(pretrained=False)
model = get_model("xception", pretrained=False)
model = nn.Sequential(*list(
model.children())[:-1]) # Remove original output layer
model[0].final_block.pool = nn.Sequential(nn.AdaptiveAvgPool2d(1)) # xcep
model = FCN(model, 2048)
model.cuda()
train_data = Task1_loader("./Task_1/train.csv", phase='train')
test_data = Task1_loader("./Task_1/test.csv", phase='test')
train_loader = DataLoader(train_data,
batch_size=batch_size,
shuffle=True,
num_workers=8)
valid_loader = DataLoader(test_data,
batch_size=batch_size,
shuffle=False,
num_workers=8)
criterion = nn.BCELoss()
# optimizer = optim.SGD(model.parameters(), lr=0.0018, momentum=0.27)
optimizer = optim.Adam(model.parameters(), lr=1e-4, weight_decay=1e-5)
train(model,
train_loader,
valid_loader,
criterion,
optimizer,
100,
device='cuda')
def main():
root_dir_train = "./Task_1/"
root_dir_test = "./Task_1/"
train_path = os.path.join(root_dir_train, "development/") # train images
test_path = os.path.join(root_dir_test, "evaluation/") # train images
create_csv(root_dir_train, root_dir_test, train_path, test_path)
preprocess = FaceRecog(margin=7)
batch_size = 55
# model = models.resnet18(pretrained=False)
model = models.resnet34(pretrained=False)
clssf = BinaryClassifier(model, freeze=False)
train_data = Task1_loader("./Task_1/train.csv",
phase='train',
preprocess=preprocess)
test_data = Task1_loader("./Task_1/test.csv",
phase='test',
preprocess=preprocess)
train_loader = DataLoader(train_data,
batch_size=batch_size,
shuffle=True,
num_workers=8)
valid_loader = DataLoader(test_data,
batch_size=batch_size,
shuffle=False,
num_workers=8)
criterion = nn.BCELoss()
# optimizer = optim.SGD(model.parameters(), lr=0.0018, momentum=0.27)
optimizer = optim.Adam(model.parameters(), lr=0.0018, weight_decay=0.0015)
train(clssf,
train_loader,
valid_loader,
criterion,
optimizer,
10,
device='cpu')
def main():
root_dir_train = "./Task_1/"
root_dir_test = "./Task_2_3/"
train_path = os.path.join(root_dir_train, "development/") # train images
test_path = os.path.join(root_dir_test, "evaluation/") # train images
create_csv(root_dir_train, root_dir_test, train_path, test_path)
preprocess = FaceRecog(margin=7)
batch_size = 35
model = models.resnet34(pretrained=False)
clssf = BinaryClassifier(model, freeze=False)
train_data = Task1_loader(root_dir_train + "train.csv",
phase='train',
preprocess=preprocess)
test_data = Task1_loader(root_dir_test + "test.csv",
phase='test',
preprocess=preprocess)
train_loader = DataLoader(train_data,
batch_size=batch_size,
shuffle=True,
num_workers=8)
valid_loader = DataLoader(test_data,
batch_size=batch_size,
shuffle=False,
num_workers=8)
criterion = nn.BCELoss()
optimizer = optim.Adam(model.parameters(), lr=0.0018, weight_decay=0.0015)
load_checkpoint(model, 'checkpoints/task1_84_7.pkl', optimizer)
valid_loss, accuracy = validate(model, valid_loader, criterion, 'cpu')
print(f'val loss: {valid_loss:04f} ' f'val acc: {valid_acc*100:.4f}%')
def main():
root_dir_train = "./Task_1/"
root_dir_test = "./Task_2_3/"
train_path = os.path.join(root_dir_train, "development/") # train images
test_path = os.path.join(root_dir_test, "evaluation/") # train images
create_csv(root_dir_train, root_dir_test, train_path, test_path)
# preprocess()
batch_size = 35
model = models.resnet50(pretrained=True)
clssf = BinaryClassifier(model)
train_data = Task1_loader(root_dir_train + "train.csv", phase='train')
test_data = Task1_loader(root_dir_test + "test.csv", phase='test')
train_loader = DataLoader(train_data,
batch_size=batch_size,
shuffle=True,
num_workers=8)
valid_loader = DataLoader(test_data,
batch_size=batch_size,
shuffle=False,
num_workers=8)
#criterion = nn.CrossEntropyLoss()
criterion = nn.BCELoss()
optimizer = optim.SGD(model.parameters(), lr=0.001, momentum=0.2)
# optimizer = optim.SGD(model.parameters(), lr=0.0015, momentum=0.26)
train(clssf,
train_loader,
valid_loader,
criterion,
optimizer,
10,
device='cpu')
model_path = os.path.join(model_dir, "scratch_resnext")
model = torch.load(model_path)
use_gpu = torch.cuda.is_available()
if use_gpu:
model = model.cuda()
batch_size = 16
epochs = 500
lr = 1e-4
momentum = 0
w_decay = 1e-5
step_size = 50
gamma = 0.5
train_data = Task1_loader("./Task_1/train.csv", phase='train')
test_data = Task1_loader("./Task_1/test.csv", phase='test')
train_loader = DataLoader(train_data,
batch_size=batch_size,
shuffle=True,
num_workers=8)
val_loader = DataLoader(train_data,
batch_size=batch_size,
shuffle=False,
num_workers=8)
activation = nn.Sigmoid()
criterion = nn.BCELoss()
optimizer = optim.RMSprop(model.parameters(),
lr=lr,
momentum=momentum,
weight_decay=w_decay)
y_true = y_targets
y_pred = y_preds
return roc_curve(y_true, y_pred)
"""
model = get_model("xception", pretrained=True)
model = nn.Sequential(*list(model.children())[:-1]) # Remove original output layer
model[0].final_block.pool = nn.Sequential(nn.AdaptiveAvgPool2d(1)) # xcep
model = FCN(model, 2048)
"""
model = torch.load("./checkpoints/89.pkl")
test_data = Task1_loader("./Task_1/test.csv", phase='test')
valid_loader = DataLoader(test_data,
batch_size=55,
shuffle=False,
num_workers=8)
y_pred = []
test_y = []
with torch.no_grad():
for batch in valid_loader:
# Move the validation batch to the GPU
inputs = Variable(batch['X'])
labels = Variable(batch['Y'])