def make_data_loader(batch_size=16):
"""
Prepare Dataset and apply Dataloader.
You don't have to change it.
"""
train_set = Dataset(split="train")
val_set = Dataset(split="val")
test_set = Dataset(split="test")
num_class = train_set.NUM_CLASSES
train_loader = DataLoader(train_set, batch_size=batch_size, shuffle=True)
val_loader = DataLoader(val_set, batch_size=batch_size, shuffle=False)
test_loader = DataLoader(test_set, batch_size=batch_size, shuffle=False)
return train_loader, val_loader, test_loader, num_class
def make_data_loader(batch_size=16, is_develop=False):
"""
Prepare Dataset and apply Dataloader.
You don't have to change it.
"""
lidc = Lidc(is_develop=is_develop)
il = lidc.img_list
ml = lidc.mask_list
train_set = Dataset(il, ml, split="train")
val_set = Dataset(il, ml, split="val")
test_set = Dataset(il, ml, split="test")
num_class = train_set.NUM_CLASSES
train_loader = DataLoader(train_set, batch_size=batch_size, shuffle=True)
val_loader = DataLoader(val_set, batch_size=batch_size, shuffle=False)
test_loader = DataLoader(test_set, batch_size=batch_size, shuffle=False)
return train_loader, val_loader, test_loader, num_class
def preprocess_data(data_config: object, img_list: List, lbl_list: List,
batch_size: int, mode: str):
"""Preprocess dataset
Parameters
----------
data_config : object
data configuration
img_list : List
a list of image paths
lbl_list : List
a list of labels
batch_size : int
batch_size
mode : str
'train' or 'eval'
Returns
-------
Object :
DataLoader instance
Raises
------
ValueError
raise value error if the mode is not 'train' or 'eval'
"""
# transform
resize = (data_config.img_size[0], data_config.img_size[1])
color_mean = tuple(data_config.color_mean)
color_std = tuple(data_config.color_std)
transform = DataTransform(resize, color_mean, color_std, mode)
# dataset
dataset = Dataset(img_list, lbl_list, transform)
# dataloader
if mode == 'train':
return data.DataLoader(dataset,
batch_size=batch_size,
shuffle=True,
num_workers=2)
elif mode == 'eval':
return data.DataLoader(dataset,
batch_size=batch_size,
shuffle=False,
num_workers=2)
else:
raise ValueError(
'the mode should be train or eval. this mode is not supported')
if __name__ == "__main__":
from model.fcos import FCOSDetector
from dataloader.dataset import Dataset
from tensorboardX import SummaryWriter
images_root = '/home'
val_path = '/mnt/hdd1/benkebishe01/data/val.txt'
transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))
])
val_loader = torch.utils.data.DataLoader(Dataset(images_root,
val_path,
img_size=512,
transform=transform,
train=False),
batch_size=64,
shuffle=False)
draw = False
if draw:
writer = SummaryWriter(comment='voc_mAP')
model = FCOSDetector(mode="inference")
# model.load_state_dict(torch.load("/mnt/hdd1/benkebishe01/FCOS/80epoch/fcos_voc_3/voc_epoch80_loss0.8893.pth"))
model = model.to(device).eval()
print("===>success loading model")
names = os.listdir(model_root)
names = sorted(names, key=functools.cmp_to_key(compare))
weight_path1 = '../src/logs/multimodal/mixup-lam_2-latest/22184_7_best_1.2906.pth'
weight_path2 = '../src/logs/multimodal/mixup-lam_2-latest-newseed/55460_19_best_1.2704.pth'
weight_path3 = '../src/logs/multimodal/mixup-lam_2-latest-newnewseed/30503_10_best_1.2922.pth'
weight_path4 = '../src/logs/multimodal/mixup-lam_2-seed1234-bigger-models/44368_7_best_1.2946.pth'
test_speech_root_dir = '../features/test/speech'
test_face_root_dir = '../features/test/video'
test_text_root_dir = '../features/test/text'
test_files = [
file for file in os.listdir(test_speech_root_dir) if '.npy' in file
]
test_files.sort()
test_dataset = Dataset(speech_root_dir=test_speech_root_dir,
video_root_dir=test_face_root_dir,
text_root_dir=test_text_root_dir,
file_list=test_files,
label_smoothing=0,
is_train=False,
is_test=True,
n_frames=32)
test_loader = data.DataLoader(dataset=test_dataset,
batch_size=32,
num_workers=20,
shuffle=False)
test3_speech_root_dir = '../features/test3/speech'
test3_face_root_dir = '../features/test3/video'
test3_text_root_dir = '../features/test3/text'
test3_files = [
file for file in os.listdir(test3_speech_root_dir) if '.npy' in file
]
test3_files.sort()
optimizer=torch.optim.Adam(model.parameters(), lr=1e-4, weight_decay=1e-4)
BATCH_SIZE = 16
EPOCHS = 60
WARMPUP_STEPS_RATIO = 0.12
transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.485,0.456,0.406), (0.229,0.224,0.225))
])
# 2012_train 2007_val
cfg= {'images_root': '/home', 'train_path': '/mnt/hdd1/benkebishe01/data/train.txt', 'test_path': '/mnt/hdd1/benkebishe01/data/val.txt',
'img_size': 512}
train_dataset = Dataset(cfg['images_root'], cfg['train_path'], img_size=cfg['img_size'], transform=transform, train=True)
train_loader = torch.utils.data.DataLoader(train_dataset, batch_size=BATCH_SIZE, shuffle=True)
test_loader = torch.utils.data.DataLoader(
Dataset(cfg['images_root'], cfg['test_path'],img_size=cfg['img_size'],
transform=transform, train=False),
batch_size=4,
shuffle=False,)
steps_per_epoch=len(train_dataset)//BATCH_SIZE
TOTAL_STEPS=steps_per_epoch*EPOCHS
WARMPUP_STEPS=TOTAL_STEPS*WARMPUP_STEPS_RATIO
# global GLOBAL_STEPS
GLOBAL_STEPS=1
LR_INIT=5e-5
transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))
])
# 2012_train 2007_val
cfg = {
'images_root': '/home',
'train_path': '/mnt/hdd1/benkebishe01/data/train.txt',
'test_path': '/mnt/hdd1/benkebishe01/data/val.txt',
'img_size': 512
}
test_loader = torch.utils.data.DataLoader(Dataset(cfg['images_root'],
cfg['test_path'],
img_size=cfg['img_size'],
transform=transform,
train=False),
batch_size=4,
shuffle=False)
draw = False
if draw:
writer = SummaryWriter(comment='test_voc_one')
# model_root = "/mnt/hdd1/benkebishe01/FCOS/fcos_giou_one_new"
model_root = "/mnt/hdd1/benkebishe01/FCOS/fcos_voc_one"
def compare(x, y):
if opt.experiment_name is not None:
print(F"Experiment name: {opt.experiment_name}")
print(F"Run directory: {opt.run_dir}")
# Pick the selected model
if opt.model == 'DeepNAG':
model_t = DeepNAG
print('Selected model: DeepNAG')
elif opt.model == 'DeepGAN':
model_t = DeepGAN
print('Selected model: DeepGAN')
else:
raise Exception(F"Unknown model '{opt.model}'")
# Determine the computation device to use
if opt.use_cuda:
print("Using CUDA")
device = torch.device('cuda:0')
else:
print("Using CPU")
device = torch.device('cpu')
# Instantiate the dataset
dataset = Dataset.instantiate(opt)
if opt.evaluate is None:
train(model_t, dataset, device)
else:
evaluate(model_t, dataset, device)
]
if not (config.mode == 'speech' or config.mode == 'multimodal'):
train_speech_root_dir = None
valid_speech_root_dir = None
if not (config.mode == 'text' or config.mode == 'multimodal'):
train_text_root_dir = None
valid_text_root_dir = None
if not (config.mode == 'face' or config.mode == 'multimodal'):
train_face_root_dir = None
valid_face_root_dir = None
train_dataset = Dataset(speech_root_dir=train_speech_root_dir,
video_root_dir=train_face_root_dir,
text_root_dir=train_text_root_dir,
file_list=train_files,
label_smoothing=config.label_smoothing,
is_train=True,
n_frames=16)
valid_dataset = Dataset(speech_root_dir=valid_speech_root_dir,
video_root_dir=valid_face_root_dir,
text_root_dir=valid_text_root_dir,
file_list=valid_files,
label_smoothing=0,
is_train=False,
n_frames=16)
train_loader = data.DataLoader(dataset=train_dataset,
batch_size=config.batch_size,
num_workers=config.num_workers,
shuffle=True)
valid_loader = data.DataLoader(dataset=valid_dataset,