def test():
if args.cuda:
print('use cuda')
cudnn.benchmark = True
device = torch.device("cuda")
else:
device = torch.device("cpu")
# load net
input_size = [args.input_size, args.input_size]
num_classes = 20
testset = VOCDetection(VOC_ROOT, img_size=None, image_sets=[('2007', 'test')], transform=None)
# build model
if args.version == 'yolo':
from models.yolo import myYOLO
net = myYOLO(device, input_size=input_size, num_classes=num_classes, trainable=False)
print('Let us test yolo on the VOC0712 dataset ......')
else:
print('Unknown Version !!!')
exit()
net.load_state_dict(torch.load(args.trained_model, map_location=device))
net.eval()
print('Finished loading model!')
net = net.to(device)
# evaluation
test_net(net, device, testset,
BaseTransform(net.input_size),
thresh=args.visual_threshold)
def test():
# get device
if args.cuda:
cudnn.benchmark = True
device = torch.device("cuda")
else:
device = torch.device("cpu")
# load net
num_classes = 80
if args.dataset == 'COCO_val':
cfg = config.coco_af
input_size = cfg['min_dim']
testset = COCODataset(data_dir=args.dataset_root,
json_file='instances_val2017.json',
name='val2017',
img_size=cfg['min_dim'][0],
debug=args.debug)
elif args.dataset == 'COCO_test-dev':
cfg = config.coco_af
input_size = cfg['min_dim']
testset = COCODataset(data_dir=args.dataset_root,
json_file='image_info_test-dev2017.json',
name='test2017',
img_size=cfg['min_dim'][0],
debug=args.debug)
elif args.dataset == 'VOC':
cfg = config.voc_af
input_size = cfg['min_dim']
testset = VOCDetection(VOC_ROOT, [('2007', 'test')], None,
VOCAnnotationTransform())
# build model
if args.version == 'yolo':
from models.yolo import myYOLO
net = myYOLO(device,
input_size=input_size,
num_classes=num_classes,
trainable=False)
print('Let us test YOLO on the %s dataset ......' % (args.dataset))
else:
print('Unknown Version !!!')
exit()
net.load_state_dict(torch.load(args.trained_model, map_location=device))
net.to(device).eval()
print('Finished loading model!')
# evaluation
test_net(net,
device,
testset,
BaseTransform(net.input_size,
mean=(0.406, 0.456, 0.485),
std=(0.225, 0.224, 0.229)),
thresh=args.visual_threshold)
def test():
if args.cuda:
print('use cuda')
cudnn.benchmark = True
device = torch.device("cuda")
else:
device = torch.device("cpu")
# load net
cfg = config.voc_af
input_size = cfg['min_dim']
num_classes = len(VOC_CLASSES)
testset = VOCDetection(args.voc_root, [('2007', 'test')], None,
VOCAnnotationTransform())
# build model
if args.version == 'yolo':
from models.yolo import myYOLO
net = myYOLO(device,
input_size=input_size,
num_classes=num_classes,
trainable=False)
print('Let us test yolo on the VOC0712 dataset ......')
else:
print('Unknown Version !!!')
exit()
net.load_state_dict(torch.load(args.trained_model, map_location=device))
net.eval()
print('Finished loading model!')
net = net.to(device)
# evaluation
test_net(net,
device,
testset,
BaseTransform(net.input_size,
mean=(0.406, 0.456, 0.485),
std=(0.225, 0.224, 0.229)),
thresh=args.visual_threshold)
if __name__ == '__main__':
if args.cuda:
print('use cuda')
torch.backends.cudnn.benchmark = True
device = torch.device("cuda")
else:
device = torch.device("cpu")
num_classes = args.num_classes
input_size = [args.input_size, args.input_size]
if args.version == 'yolo':
from models.yolo import myYOLO
model = myYOLO(device,
input_size=input_size,
num_classes=num_classes,
trainable=False)
print('Let us evaluate YOLO on the COCO dataset ......')
else:
print('Unknown Version !!!')
exit()
# load model
model.load_state_dict(torch.load(args.trained_model, map_location=device))
model.eval().to(device)
print('Finished loading model!')
test(model, device, input_size)
def train():
args = parse_args()
path_to_save = os.path.join(args.save_folder, args.version)
os.makedirs(path_to_save, exist_ok=True)
hr = False
if args.high_resolution:
print('use hi-res backbone')
hr = True
cfg = voc_af
if args.cuda:
print('use cuda')
cudnn.benchmark = True
device = torch.device("cuda")
else:
device = torch.device("cpu")
# use multi-scale trick
if args.multi_scale:
print('use multi-scale trick.')
input_size = [608, 608]
dataset = VOCDetection(root=args.dataset_root,
transform=SSDAugmentation([608, 608],
mean=(0.406, 0.456,
0.485),
std=(0.225, 0.224,
0.229)))
else:
input_size = cfg['min_dim']
dataset = VOCDetection(root=args.dataset_root,
transform=SSDAugmentation(cfg['min_dim'],
mean=(0.406, 0.456,
0.485),
std=(0.225, 0.224,
0.229)))
# build model
if args.version == 'yolo':
from models.yolo import myYOLO
yolo_net = myYOLO(device,
input_size=input_size,
num_classes=args.num_classes,
trainable=True,
hr=hr)
print('Let us train yolo on the VOC0712 dataset ......')
else:
print('We only support YOLO !!!')
exit()
# finetune the model trained on COCO
if args.resume is not None:
print('finetune COCO trained ')
yolo_net.load_state_dict(torch.load(args.resume, map_location=device),
strict=False)
# use tfboard
if args.tfboard:
print('use tensorboard')
from torch.utils.tensorboard import SummaryWriter
c_time = time.strftime('%Y-%m-%d %H:%M:%S',
time.localtime(time.time()))
log_path = os.path.join('log/voc/', args.version, c_time)
os.makedirs(log_path, exist_ok=True)
writer = SummaryWriter(log_path)
print(
"----------------------------------------Object Detection--------------------------------------------"
)
model = yolo_net
model.to(device)
base_lr = args.lr
tmp_lr = base_lr
optimizer = optim.SGD(model.parameters(),
lr=args.lr,
momentum=args.momentum,
weight_decay=args.weight_decay)
# loss counters
print("----------------------------------------------------------")
print("Let's train OD network !")
print('Training on:', dataset.name)
print('The dataset size:', len(dataset))
print("----------------------------------------------------------")
epoch_size = len(dataset) // args.batch_size
max_epoch = cfg['max_epoch']
data_loader = data.DataLoader(dataset,
args.batch_size,
num_workers=args.num_workers,
shuffle=True,
collate_fn=detection_collate,
pin_memory=True)
# create batch iterator
t0 = time.time()
# start training
for epoch in range(max_epoch):
# use cos lr
if args.cos and epoch > 20 and epoch <= max_epoch - 20:
# use cos lr
tmp_lr = 0.00001 + 0.5 * (base_lr - 0.00001) * (
1 + math.cos(math.pi * (epoch - 20) * 1. / (max_epoch - 20)))
set_lr(optimizer, tmp_lr)
elif args.cos and epoch > max_epoch - 20:
tmp_lr = 0.00001
set_lr(optimizer, tmp_lr)
# use step lr
else:
if epoch in cfg['lr_epoch']:
tmp_lr = tmp_lr * 0.1
set_lr(optimizer, tmp_lr)
for iter_i, (images, targets) in enumerate(data_loader):
# WarmUp strategy for learning rate
if not args.no_warm_up:
if epoch < args.wp_epoch:
tmp_lr = base_lr * pow((iter_i + epoch * epoch_size) * 1. /
(args.wp_epoch * epoch_size), 4)
# tmp_lr = 1e-6 + (base_lr-1e-6) * (iter_i+epoch*epoch_size) / (epoch_size * (args.wp_epoch))
set_lr(optimizer, tmp_lr)
elif epoch == args.wp_epoch and iter_i == 0:
tmp_lr = base_lr
set_lr(optimizer, tmp_lr)
targets = [label.tolist() for label in targets]
# make train label
targets = tools.gt_creator(input_size=input_size,
stride=yolo_net.stride,
label_lists=targets)
# to device
images = images.to(device)
targets = torch.tensor(targets).float().to(device)
# forward and loss
conf_loss, cls_loss, txtytwth_loss, total_loss = model(
images, target=targets)
# backprop and update
total_loss.backward()
optimizer.step()
optimizer.zero_grad()
if iter_i % 10 == 0:
if args.tfboard:
# viz loss
writer.add_scalar('object loss', conf_loss.item(),
iter_i + epoch * epoch_size)
writer.add_scalar('class loss', cls_loss.item(),
iter_i + epoch * epoch_size)
writer.add_scalar('local loss', txtytwth_loss.item(),
iter_i + epoch * epoch_size)
t1 = time.time()
print(
'[Epoch %d/%d][Iter %d/%d][lr %.6f]'
'[Loss: obj %.2f || cls %.2f || bbox %.2f || total %.2f || size %d || time: %.2f]'
% (epoch + 1, max_epoch, iter_i, epoch_size, tmp_lr,
conf_loss.item(), cls_loss.item(), txtytwth_loss.item(),
total_loss.item(), input_size[0], t1 - t0),
flush=True)
t0 = time.time()
# multi-scale trick
if iter_i % 10 == 0 and iter_i > 0 and args.multi_scale:
size = random.randint(10, 19) * 32
input_size = [size, size]
# change input dim
# But this operation will report bugs when we use more workers in data loader, so I have to use 0 workers.
# I don't know how to make it suit more workers, and I'm trying to solve this question.
data_loader.dataset.reset_transform(
SSDAugmentation(input_size,
mean=(0.406, 0.456, 0.485),
std=(0.225, 0.224, 0.229)))
if (epoch + 1) % 10 == 0:
print('Saving state, epoch:', epoch + 1)
torch.save(
model.state_dict(),
os.path.join(path_to_save,
args.version + '_' + repr(epoch + 1) + '.pth'))
if __name__ == '__main__':
global cfg
cfg = coco_af
if args.cuda:
print('use cuda')
torch.backends.cudnn.benchmark = True
device = torch.device("cuda")
else:
device = torch.device("cpu")
num_classes = args.num_classes
if args.version == 'yolo':
from models.yolo import myYOLO
model = myYOLO(device,
input_size=cfg['min_dim'],
num_classes=num_classes,
trainable=False)
print('Let us evaluate YOLO on the COCO dataset ......')
else:
print('Unknown Version !!!')
exit()
# load model
model.load_state_dict(torch.load(args.trained_model, map_location=device))
model.eval().to(device)
print('Finished loading model!')
test(model, device)
def train():
args = parse_args()
path_to_save = os.path.join(args.save_folder, args.dataset, args.version)
# path_to_save = os.path.join('weights', 'voc', 'yolov')
os.makedirs(path_to_save, exist_ok=True)
# use hi-res backbone
if args.high_resolution:
print('use hi-res backbone')
hr = True
else:
hr = False
# cuda
if args.cuda:
print('use cuda')
cudnn.benchmark = True
device = torch.device("cuda", 2)
else:
device = torch.device("cpu")
# multi-scale
if args.multi_scale:
print('use the multi-scale trick ...')
train_size = [640, 640]
val_size = [416, 416]
else:
train_size = [416, 416]
val_size = [416, 416]
cfg = train_cfg
# dataset and evaluator
print("Setting Arguments.. : ", args)
print("----------------------------------------------------------")
print('Loading the dataset...')
if args.dataset == 'voc':
VOC_ROOT = '/home/liyi219/python/new-YOLOv1_PyTorch-master/VOCdevkit'
data_dir = VOC_ROOT
num_classes = 20
dataset = VOCDetection(root=data_dir,
img_size=train_size[0],
transform=SSDAugmentation(train_size))
evaluator = VOCAPIEvaluator(data_root=data_dir,
img_size=val_size,
device=device,
transform=BaseTransform(val_size),
labelmap=VOC_CLASSES)
# elif args.dataset == 'coco':
# data_dir = coco_root
# num_classes = 80
# dataset = COCODataset(
# data_dir=data_dir,
# img_size=train_size[0],
# transform=SSDAugmentation(train_size),
# debug=args.debug
# )
#
# evaluator = COCOAPIEvaluator(
# data_dir=data_dir,
# img_size=val_size,
# device=device,
# transform=BaseTransform(val_size)
# )
else:
print('unknow dataset !! Only support voc and coco !!')
exit(0)
print('Training model on:', dataset.name)
print('The dataset size:', len(dataset))
print("----------------------------------------------------------")
# dataloader
dataloader = torch.utils.data.DataLoader(dataset,
batch_size=args.batch_size,
shuffle=True,
collate_fn=detection_collate,
num_workers=args.num_workers,
pin_memory=True)
"""
dataloader = torch.utils.data.DataLoader(
dataset,
batch_size=32,
shuffle=True,
collate_fn=detection_collate,
num_workers=8,
pin_memory=True
)
"""
# build model
if args.version == 'yolo':
from models.yolo import myYOLO
yolo_net = myYOLO(device,
input_size=train_size,
num_classes=num_classes,
trainable=True)
print('Let us train yolo on the %s dataset ......' % (args.dataset))
else:
print('We only support YOLO !!!')
exit()
model = yolo_net
model.to(device).train()
# use tfboard
if args.tfboard:
print('use tensorboard')
from torch.utils.tensorboard import SummaryWriter
c_time = time.strftime('%Y-%m-%d %H:%M:%S',
time.localtime(time.time()))
log_path = os.path.join('log/coco/', args.version, c_time)
# log_path = os.path.join('log/coco/', 'yolo', c_time)
os.makedirs(log_path, exist_ok=True)
writer = SummaryWriter(log_path)
# keep training
if args.resume is not None:
print('keep training model: %s' % (args.resume))
model.load_state_dict(torch.load(args.resume, map_location=device))
# model.load_state_dict(torch.load('keep training', map_location=device))
# optimizer setup
base_lr = args.lr
tmp_lr = base_lr
optimizer = optim.SGD(model.parameters(),
lr=args.lr,
momentum=args.momentum,
weight_decay=args.weight_decay)
"""
optimizer = optim.SGD(model.parameters(),
lr=1e-3,
momentum=0.9,
weight_decay=5e-4
)
"""
max_epoch = cfg['max_epoch']
epoch_size = len(dataset) // args.batch_size
# epoch_size = len(dataset) // 32
# start training loop
t0 = time.time()
for epoch in range(args.start_epoch, max_epoch):
# use cos lr
if args.cos and epoch > 20 and epoch <= max_epoch - 20:
# use cos lr
tmp_lr = 0.00001 + 0.5 * (base_lr - 0.00001) * (
1 + math.cos(math.pi * (epoch - 20) * 1. / (max_epoch - 20)))
set_lr(optimizer, tmp_lr)
elif args.cos and epoch > max_epoch - 20:
tmp_lr = 0.00001
set_lr(optimizer, tmp_lr)
# use step lr
else:
if epoch in cfg['lr_epoch']:
tmp_lr = tmp_lr * 0.1
set_lr(optimizer, tmp_lr)
for iter_i, (images, targets) in enumerate(dataloader):
# WarmUp strategy for learning rate
if not args.no_warm_up:
if epoch < args.wp_epoch:
tmp_lr = base_lr * pow((iter_i + epoch * epoch_size) * 1. /
(args.wp_epoch * epoch_size), 4)
# tmp_lr = base_lr * pow((iter_i + epoch * epoch_size) * 1. / (2 * epoch_size), 4)
# tmp_lr = 1e-6 + (base_lr-1e-6) * (iter_i+epoch*epoch_size) / (epoch_size * (args.wp_epoch))
set_lr(optimizer, tmp_lr)
elif epoch == args.wp_epoch and iter_i == 0:
tmp_lr = base_lr
set_lr(optimizer, tmp_lr)
# to device
images = images.to(device)
# multi-scale trick
if iter_i % 10 == 0 and iter_i > 0 and args.multi_scale:
# randomly choose a new size
size = random.randint(10, 19) * 32
train_size = [size, size]
model.set_grid(train_size)
if args.multi_scale:
# interpolate
images = torch.nn.functional.interpolate(images,
size=train_size,
mode='bilinear',
align_corners=False)
# make train label
# 原本的targets里面只有bbox坐标和类别参数,用gt_creator函数加上置信度和权重参数
targets = [label.tolist() for label in targets]
targets = tools.gt_creator(input_size=train_size,
stride=yolo_net.stride,
label_lists=targets)
targets = torch.tensor(targets).float().to(device)
# forward and loss
conf_loss, cls_loss, txtytwth_loss, total_loss = model(
images, target=targets)
# backprop
total_loss.backward()
optimizer.step()
optimizer.zero_grad()
# display
if iter_i % 10 == 0:
if args.tfboard:
# viz loss
writer.add_scalar('object loss', conf_loss.item(),
iter_i + epoch * epoch_size)
writer.add_scalar('class loss', cls_loss.item(),
iter_i + epoch * epoch_size)
writer.add_scalar('local loss', txtytwth_loss.item(),
iter_i + epoch * epoch_size)
t1 = time.time()
print(
'[Epoch %d/%d][Iter %d/%d][lr %.6f]'
'[Loss: obj %.2f || cls %.2f || bbox %.2f || total %.2f || size %d || time: %.2f]'
% (epoch + 1, max_epoch, iter_i, epoch_size, tmp_lr,
conf_loss.item(), cls_loss.item(), txtytwth_loss.item(),
total_loss.item(), train_size[0], t1 - t0),
flush=True)
t0 = time.time()
# evaluation
if (epoch + 1) % args.eval_epoch == 0:
model.trainable = False
model.set_grid(val_size)
model.eval()
# evaluate
evaluator.evaluate(model)
# convert to training mode.
model.trainable = True
model.set_grid(train_size)
model.train()
# save model
if (epoch + 1) % 10 == 0:
print('Saving state, epoch:', epoch + 1)
torch.save(
model.state_dict(),
os.path.join(path_to_save,
args.version + '_' + repr(epoch + 1) + '.pth'))
def train():
args = parse_args()
data_dir = coco_root
path_to_save = os.path.join(args.save_folder, args.version)
os.makedirs(path_to_save, exist_ok=True)
hr = False
if args.high_resolution:
print('use hi-res backbone')
hr = True
cfg = coco_af
if args.cuda:
print('use cuda')
cudnn.benchmark = True
device = torch.device("cuda")
else:
device = torch.device("cpu")
if args.mosaic:
print("use Mosaic Augmentation ...")
# multi scale
if args.multi_scale:
print('Let us use the multi-scale trick.')
input_size = [640, 640]
else:
input_size = [416, 416]
print("Setting Arguments.. : ", args)
print("----------------------------------------------------------")
print('Loading the MSCOCO dataset...')
# dataset
dataset = COCODataset(data_dir=data_dir,
img_size=input_size[0],
transform=SSDAugmentation(input_size),
debug=args.debug,
mosaic=args.mosaic)
# build model
if args.version == 'yolo':
from models.yolo import myYOLO
yolo_net = myYOLO(device,
input_size=input_size,
num_classes=args.num_classes,
trainable=True,
hr=hr)
print('Let us train yolo on the COCO dataset ......')
else:
print('We only support YOLO !!!')
exit()
print("----------------------------------------------------------")
print('The dataset size:', len(dataset))
print("----------------------------------------------------------")
# use tfboard
if args.tfboard:
print('use tensorboard')
from torch.utils.tensorboard import SummaryWriter
c_time = time.strftime('%Y-%m-%d %H:%M:%S',
time.localtime(time.time()))
log_path = os.path.join('log/coco/', args.version, c_time)
os.makedirs(log_path, exist_ok=True)
writer = SummaryWriter(log_path)
model = yolo_net
model.to(device).train()
dataloader = torch.utils.data.DataLoader(dataset,
batch_size=args.batch_size,
shuffle=True,
collate_fn=detection_collate,
num_workers=args.num_workers)
evaluator = COCOAPIEvaluator(data_dir=data_dir,
img_size=cfg['min_dim'],
device=device,
transform=BaseTransform(cfg['min_dim']))
# optimizer setup
base_lr = args.lr
tmp_lr = base_lr
optimizer = optim.SGD(model.parameters(),
lr=args.lr,
momentum=args.momentum,
weight_decay=args.weight_decay)
max_epoch = cfg['max_epoch']
epoch_size = len(dataset) // args.batch_size
# start training loop
t0 = time.time()
for epoch in range(max_epoch):
# use cos lr
if args.cos and epoch > 20 and epoch <= max_epoch - 20:
# use cos lr
tmp_lr = 0.00001 + 0.5 * (base_lr - 0.00001) * (
1 + math.cos(math.pi * (epoch - 20) * 1. / (max_epoch - 20)))
set_lr(optimizer, tmp_lr)
elif args.cos and epoch > max_epoch - 20:
tmp_lr = 0.00001
set_lr(optimizer, tmp_lr)
# use step lr
else:
if epoch in cfg['lr_epoch']:
tmp_lr = tmp_lr * 0.1
set_lr(optimizer, tmp_lr)
for iter_i, (images, targets) in enumerate(dataloader):
# WarmUp strategy for learning rate
if not args.no_warm_up:
if epoch < args.wp_epoch:
tmp_lr = base_lr * pow((iter_i + epoch * epoch_size) * 1. /
(args.wp_epoch * epoch_size), 4)
# tmp_lr = 1e-6 + (base_lr-1e-6) * (iter_i+epoch*epoch_size) / (epoch_size * (args.wp_epoch))
set_lr(optimizer, tmp_lr)
elif epoch == args.wp_epoch and iter_i == 0:
tmp_lr = base_lr
set_lr(optimizer, tmp_lr)
# to device
images = images.to(device)
# multi-scale trick
if iter_i % 10 == 0 and iter_i > 0 and args.multi_scale:
# randomly choose a new size
size = random.randint(10, 19) * 32
input_size = [size, size]
model.set_grid(input_size)
if args.multi_scale:
# interpolate
images = torch.nn.functional.interpolate(images,
size=input_size,
mode='bilinear',
align_corners=False)
# make labels
targets = [label.tolist() for label in targets]
targets = tools.gt_creator(input_size=input_size,
stride=yolo_net.stride,
label_lists=targets)
targets = torch.tensor(targets).float().to(device)
# forward and loss
conf_loss, cls_loss, txtytwth_loss, total_loss = model(
images, target=targets)
# backprop
total_loss.backward()
optimizer.step()
optimizer.zero_grad()
if args.tfboard:
# viz loss
writer.add_scalar('object loss', conf_loss.item(),
iter_i + epoch * epoch_size)
writer.add_scalar('class loss', cls_loss.item(),
iter_i + epoch * epoch_size)
writer.add_scalar('local loss', txtytwth_loss.item(),
iter_i + epoch * epoch_size)
if iter_i % 10 == 0:
t1 = time.time()
print(
'[Epoch %d/%d][Iter %d/%d][lr %.6f]'
'[Loss: obj %.2f || cls %.2f || bbox %.2f || total %.2f || size %d || time: %.2f]'
% (epoch + 1, max_epoch, iter_i, epoch_size, tmp_lr,
conf_loss.item(), cls_loss.item(), txtytwth_loss.item(),
total_loss.item(), input_size[0], t1 - t0),
flush=True)
t0 = time.time()
if (epoch + 1) % 10 == 0:
print('Saving state, epoch:', epoch + 1)
torch.save(
model.state_dict(),
os.path.join(path_to_save,
args.version + '_' + repr(epoch + 1) + '.pth'))
# COCO evaluation
if (epoch + 1) % args.eval_epoch == 0:
model.trainable = False
model.set_grid(cfg['min_dim'])
# evaluate
ap50_95, ap50 = evaluator.evaluate(model)
print('ap50 : ', ap50)
print('ap50_95 : ', ap50_95)
# convert to training mode.
model.trainable = True
model.train()
if args.tfboard:
writer.add_scalar('val/COCOAP50', ap50, epoch + 1)
writer.add_scalar('val/COCOAP50_95', ap50_95, epoch + 1)
