def _tranform(self):
train_z_transforms = transforms.Compose([ToTensor()])
train_x_transforms = transforms.Compose([ToTensor()])
self.ret['train_x_transforms'] = train_x_transforms(
self.ret['instance_img'])
self.ret['train_z_transforms'] = train_z_transforms(
self.ret['exemplar_img'])
def __init__(self, params, model_path = None, name='SiamRPN', **kargs):
super(TrackerSiamRPNLate, self).__init__(name=name, is_deterministic=True)
self.model = SiameseAlexNetLate()
self.cuda = torch.cuda.is_available()
self.device = torch.device('cuda:0' if self.cuda else 'cpu')
checkpoint = torch.load(model_path, map_location = self.device)
#print("1")
if 'model' in checkpoint.keys():
self.model.load_state_dict(torch.load(model_path, map_location = self.device)['model'])
else:
self.model.load_state_dict(torch.load(model_path, map_location = self.device))
if self.cuda:
self.model = self.model.cuda()
self.model.eval()
self.transforms = transforms.Compose([
ToTensor()
])
valid_scope = 2 * config.valid_scope + 1
self.anchors = util.generate_anchors( config.total_stride,
config.anchor_base_size,
config.anchor_scales,
config.anchor_ratios,
valid_scope)
self.window = np.tile(np.outer(np.hanning(config.score_size), np.hanning(config.score_size))[None, :],
[config.anchor_num, 1, 1]).flatten()
self.data_loader = TrackerRGBTDataLoader()
self.old_loader = TrackerDataLoader()
def __init__(self, model_path, gpu_id):
self.gpu_id = gpu_id
with torch.cuda.device(gpu_id):
self.model = SiamFCNet(training=False)
self.model.load_state_dict(torch.load(model_path))
self.model = self.model.cuda()
self.model.eval()
self.transforms = transforms.Compose([ToTensor()])
def __init__(self, model_path, gpu_id):
#print(model_path)
self.gpu_id = gpu_id
with torch.cuda.device(gpu_id):
self.model = SiameseNet()
# Since we created the model using nn.DataParallel, we use it
# again to load the state_dict
self.model.features = torch.nn.DataParallel(self.model.features)
self.model.load_state_dict(torch.load(model_path))
self.model = self.model.cuda()
self.model.eval()
self.transforms = transforms.Compose([ToTensor()])
def __init__(self, model_path, gpu_id):
self.gpu_id = gpu_id
with torch.cuda.device(gpu_id):
self.model = SiamRPN()
self.model.load_model(model_path)
self.model = self.model.cuda()
self.model.eval()
self.response_sz = config.response_sz
self.anchors = generate_anchors(config.total_stride,
config.anchor_base_size,
config.anchor_scales,
config.anchor_ratios, self.response_sz)
self.transforms = transforms.Compose([ToTensor()])
def main():
global args, best_prec1
args = parser.parse_args()
loss_list = []
acc_list = []
# create Experiment directories
cwd = os.getcwd()
siamfc = os.path.join(cwd, 'siamfc')
training_exp = os.path.join(cwd, 'training_exp')
time = datetime.now().strftime('%d-%m-%Y_%H-%M-%S')
dataset_name = args.datadir.split('/')[-1]
dataset_exp_dir = os.path.join(training_exp, dataset_name)
new_exp_dir = os.path.join(dataset_exp_dir, time)
tensorboard_dir = os.path.join(new_exp_dir + '/tensorboard/')
models_dir = os.path.join(new_exp_dir + '/models/')
output_files = [os.path.join(cwd, 'train_siamfc.sh')]
if not os.path.exists(new_exp_dir):
os.makedirs(tensorboard_dir, exist_ok=True)
os.makedirs(models_dir, exist_ok=True)
print('New experiment folder created')
for file in glob.glob(os.path.join(siamfc, "*.py")):
if os.path.isfile(file):
shutil.copyfile(file, new_exp_dir + file.split(siamfc)[1])
for file in glob.glob(os.path.join(cwd, "*.out")):
output_files.append(file)
print('\n================= EXPERIMENT START TIME', time,
'=================\n')
# Create Tensorboard summary writer
writer = SummaryWriter(tensorboard_dir)
# loading meta data
meta_data_path = os.path.join(args.datadir, "meta_data.pkl")
meta_data = pickle.load(open(meta_data_path, 'rb'))
videos = [x[0] for x in meta_data]
print('Training with', args.datadir.split('/')[-1])
# split train/valid dataset
train_videos, valid_videos = train_test_split(videos,
test_size=1 -
config.train_ratio)
# define transforms
random_crop_size = config.instance_size - 2 * config.total_stride
train_reference_transforms = transforms.Compose([
CenterCrop((config.exemplar_size, config.exemplar_size)),
Normalize(),
ToTensor()
])
train_search_transforms = transforms.Compose([
RandomCrop((random_crop_size, random_crop_size), config.max_translate),
Normalize(),
ToTensor()
])
valid_reference_transforms = transforms.Compose([
CenterCrop((config.exemplar_size, config.exemplar_size)),
Normalize(),
ToTensor()
])
valid_search_transforms = transforms.Compose([Normalize(), ToTensor()])
# opem lmdb
db = lmdb.open(args.datadir + '.lmdb', readonly=True, map_size=int(50e9))
# create dataset
train_dataset = ImagnetVIDDataset(db, train_videos, args.datadir,
train_reference_transforms,
train_search_transforms)
valid_dataset = ImagnetVIDDataset(db,
valid_videos,
args.datadir,
valid_reference_transforms,
valid_search_transforms,
training=False)
# create dataloader
print('Loading Train Dataset...')
trainloader = DataLoader(train_dataset,
batch_size=config.train_batch_size,
shuffle=True,
pin_memory=True,
num_workers=config.train_num_workers,
drop_last=True)
print('Loading Validation Dataset...')
validloader = DataLoader(valid_dataset,
batch_size=config.valid_batch_size,
shuffle=False,
pin_memory=True,
num_workers=config.valid_num_workers,
drop_last=True)
print('Initializing SiameseNet with {} Loss...'.format(args.loss.upper()))
model = siamnet.SiameseNet(loss=args.loss)
model.features = torch.nn.DataParallel(model.features)
model.init_weights()
model = model.cuda()
print("Available GPUs:", torch.cuda.device_count())
print("Model running on GPU:", next(model.parameters()).is_cuda), '\n\n'
cudnn.benchmark = True
optimizer = torch.optim.SGD(model.parameters(),
lr=config.lr,
momentum=config.momentum,
weight_decay=config.weight_decay)
scheduler = StepLR(optimizer,
step_size=config.step_size,
gamma=config.gamma)
for epoch in range(config.start_epoch, config.epoch):
# model.train() tells your model that you are training the model.
# So effectively layers like dropout, batchnorm etc. which behave
# different on the train and test procedures
training_loss = []
model.train()
for i, data in enumerate(tqdm(trainloader)):
reference_imgs, search_imgs = data
# Variable is a thin wrapper around a Tensor object,
# that also holds the gradient w.r.t. to it.
reference_var = Variable(reference_imgs).cuda()
search_var = Variable(search_imgs).cuda()
# we need to set the gradients to zero before starting
# to do backpropragation because PyTorch accumulates
# the gradients on subsequent backward passes.
optimizer.zero_grad()
outputs = model(reference_var, search_var)
loss = model.compute_loss(outputs)
loss.backward()
optimizer.step()
step = epoch * len(trainloader) + i
writer.add_scalars('Loss', {'Training': loss.data}, step)
training_loss.append(loss.data)
training_loss = torch.mean(torch.stack(training_loss)).item()
valid_loss = []
model.eval()
for i, data in enumerate(tqdm(validloader)):
reference_imgs, search_imgs = data
reference_var = Variable(reference_imgs.cuda())
search_var = Variable(search_imgs.cuda())
outputs = model(reference_var, search_var)
loss = model.compute_loss(outputs)
valid_loss.append(loss.data)
valid_loss = torch.mean(torch.stack(valid_loss)).item()
print("EPOCH %d Training Loss: %.4f, Validation Loss: %.4f" %
(epoch, training_loss, valid_loss))
torch.save(model.cpu().state_dict(),
models_dir + "siamfc_{}.pth".format(epoch + 1))
writer.add_scalars('Loss', {'Validation': valid_loss},
(epoch + 1) * len(trainloader))
model.cuda()
scheduler.step()
time = datetime.now().strftime('%d-%m-%Y-%H:%M:%S')
print('\n================= EXPERIMENT END TIME', time,
'=================\n')
# Copy slurm output files to experiment folder
for file in output_files:
try:
shutil.copyfile(file, new_exp_dir + file.split(cwd)[1])
except Exception as error:
print('Error copying file {} to experiment folder: {}'.format(
file, error))
pass
def main():
'''parameter initialization'''
args = parser.parse_args()
exp_name_dir = util.experiment_name_dir(args.experiment_name)
'''model on gpu'''
model = TrackerSiamRPN()
'''setup train data loader'''
name = 'GOT-10k'
assert name in ['VID', 'GOT-10k', 'All', 'RGBT-234']
if name == 'GOT-10k':
root_dir_RGBT234 = args.train_path
root_dir_GTOT = '/home/krautsct/Grayscale-Thermal-Dataset'
seq_dataset_rgb = GOT10k(root_dir_RGBT234, subset='train_i')
seq_dataset_i = GOT10k(root_dir_RGBT234,
subset='train_i',
visible=False)
elif name == 'VID':
root_dir = '/home/arbi/desktop/ILSVRC'
seq_dataset = ImageNetVID(root_dir, subset=('train'))
elif name == 'All':
root_dir_vid = '/home/arbi/desktop/ILSVRC'
seq_datasetVID = ImageNetVID(root_dir_vid, subset=('train'))
root_dir_got = args.train_path
seq_datasetGOT = GOT10k(root_dir_got, subset='train')
seq_dataset = util.data_split(seq_datasetVID, seq_datasetGOT)
elif name == 'RGBT-234':
root_dir = args.train_path
seq_dataset = RGBTSequence(root_dir, subset='train')
seq_dataset_val = RGBTSequence(root_dir, subset='val')
print('seq_dataset', len(seq_dataset_rgb))
train_z_transforms = transforms.Compose([ToTensor()])
train_x_transforms = transforms.Compose([ToTensor()])
train_data_ir = TrainDataLoader_ir(seq_dataset_i, train_z_transforms,
train_x_transforms, name)
anchors = train_data_ir.anchors
train_loader_ir = DataLoader(dataset=train_data_ir,
batch_size=config.train_batch_size,
shuffle=True,
num_workers=config.train_num_workers,
pin_memory=True)
train_data_rgb = TrainDataLoader(seq_dataset_rgb, train_z_transforms,
train_x_transforms, name)
anchors = train_data_rgb.anchors
train_loader_rgb = DataLoader(dataset=train_data_rgb,
batch_size=config.train_batch_size,
shuffle=True,
num_workers=config.train_num_workers,
pin_memory=True)
'''setup val data loader'''
name = 'GOT-10k'
assert name in ['VID', 'GOT-10k', 'All', 'RGBT-234']
if name == 'GOT-10k':
val_dir = '/home/krautsct/RGB-t-Val'
seq_dataset_val_rgb = GOT10k(val_dir, subset='train_i')
seq_dataset_val_ir = GOT10k(val_dir, subset='train_i', visible=False)
elif name == 'VID':
root_dir = '/home/arbi/desktop/ILSVRC'
seq_dataset_val = ImageNetVID(root_dir, subset=('val'))
elif name == 'All':
root_dir_vid = '/home/arbi/desktop/ILSVRC'
seq_datasetVID = ImageNetVID(root_dir_vid, subset=('val'))
root_dir_got = args.train_path
seq_datasetGOT = GOT10k(root_dir_got, subset='val')
seq_dataset_val = util.data_split(seq_datasetVID, seq_datasetGOT)
print('seq_dataset_val', len(seq_dataset_val_rgb))
valid_z_transforms = transforms.Compose([ToTensor()])
valid_x_transforms = transforms.Compose([ToTensor()])
val_data = TrainDataLoader_ir(seq_dataset_val_ir, valid_z_transforms,
valid_x_transforms, name)
val_loader_ir = DataLoader(dataset=val_data,
batch_size=config.valid_batch_size,
shuffle=False,
num_workers=config.valid_num_workers,
pin_memory=True)
val_data_rgb = TrainDataLoader(seq_dataset_val_rgb, valid_z_transforms,
valid_x_transforms, name)
val_loader_rgb = DataLoader(dataset=val_data_rgb,
batch_size=config.valid_batch_size,
shuffle=False,
num_workers=config.valid_num_workers,
pin_memory=True)
val_losslist = []
'''load weights'''
if not args.checkpoint_path == None:
assert os.path.isfile(
args.checkpoint_path), '{} is not valid checkpoint_path'.format(
args.checkpoint_path)
checkpoint = torch.load(args.checkpoint_path, map_location='cpu')
if 'model' in checkpoint.keys():
model.net.load_state_dict(
torch.load(args.checkpoint_path, map_location='cpu')['model'])
else:
model.net.load_state_dict(
torch.load(args.checkpoint_path, map_location='cpu'))
torch.cuda.empty_cache()
print('You are loading the model.load_state_dict')
elif config.pretrained_model:
checkpoint = torch.load(config.pretrained_model)
# change name and load parameters
checkpoint = {
k.replace('features.features', 'featureExtract'): v
for k, v in checkpoint.items()
}
model_dict = model.net.state_dict()
model_dict.update(checkpoint)
model.net.load_state_dict(model_dict)
#torch.cuda.empty_cache()
'''train phase'''
train_closses, train_rlosses, train_tlosses = AverageMeter(), AverageMeter(
), AverageMeter()
val_closses, val_rlosses, val_tlosses = AverageMeter(), AverageMeter(
), AverageMeter()
#train_val_plot = SavePlot(exp_name_dir, 'train_val_plot')
val_plot = SavePlotVal(exp_name_dir, 'val_plot')
for epoch in range(config.epoches):
model.net.train()
if config.fix_former_3_layers:
util.freeze_layers(model.net)
print('Train epoch {}/{}'.format(epoch + 1, config.epoches))
train_loss = []
with tqdm(total=config.train_epoch_size) as progbar:
for i, (dataset_rgb, dataset_ir) in enumerate(
zip(train_loader_rgb, train_loader_ir)):
#for i, dataset_rgb in enumerate(train_loader_rgb):
closs, rloss, loss = model.step(epoch,
dataset_rgb,
dataset_ir,
anchors,
epoch,
i,
train=True)
closs_ = closs.cpu().item()
if np.isnan(closs_):
sys.exit(0)
train_closses.update(closs.cpu().item())
train_rlosses.update(rloss.cpu().item())
train_tlosses.update(loss.cpu().item())
progbar.set_postfix(closs='{:05.3f}'.format(train_closses.avg),
rloss='{:05.5f}'.format(train_rlosses.avg),
tloss='{:05.3f}'.format(train_tlosses.avg))
progbar.update()
train_loss.append(train_tlosses.avg)
if i >= config.train_epoch_size - 1:
'''save model'''
model.save(model, exp_name_dir, epoch)
break
train_loss = np.mean(train_loss)
'''val phase'''
val_loss = []
with tqdm(total=config.val_epoch_size) as progbar:
print('Val epoch {}/{}'.format(epoch + 1, config.epoches))
for i, (dataset_rgb,
dataset_ir) in enumerate(zip(val_loader_rgb,
val_loader_ir)):
#for i, dataset_rgb in enumerate(val_loader_rgb):
val_closs, val_rloss, val_tloss = model.step(epoch,
dataset_rgb,
dataset_ir,
anchors,
epoch,
train=False)
closs_ = val_closs.cpu().item()
if np.isnan(closs_):
sys.exit(0)
val_closses.update(val_closs.cpu().item())
val_rlosses.update(val_rloss.cpu().item())
val_tlosses.update(val_tloss.cpu().item())
progbar.set_postfix(closs='{:05.3f}'.format(val_closses.avg),
rloss='{:05.5f}'.format(val_rlosses.avg),
tloss='{:05.3f}'.format(val_tlosses.avg))
progbar.update()
val_loss.append(val_tlosses.avg)
if i >= config.val_epoch_size - 1:
break
val_loss = np.mean(val_loss)
#train_val_plot.update(train_loss, val_loss)
val_plot.update(val_loss)
val_losslist.append(val_loss)
print('Train loss: {}, val loss: {}'.format(train_loss, val_loss))
record_path = os.path.dirname(exp_name_dir)
if not os.path.isdir(record_path):
os.makedirs(record_path)
record_file = os.path.join(exp_name_dir, 'val_losses.txt')
np.savetxt(record_file, val_losslist, fmt='%.3f', delimiter=',')
sunset_model = torch.nn.DataParallel(sunset_model)
sunrise_model.cuda()
sunset_model.cuda()
sunrise_model.eval()
sunset_model.eval()
data = WebcamData()
days = data.days
if constants.CENTER:
test_transformations = torchvision.transforms.Compose(
[Resize(),
RandomPatch(constants.PATCH_SIZE),
Center(),
ToTensor()])
else:
test_transformations = torchvision.transforms.Compose(
[Resize(), RandomPatch(constants.PATCH_SIZE),
ToTensor()])
test_dataset = Test(data, test_transformations)
#test_dataset.set_mode('sunrise')
if torch.cuda.is_available():
pin_memory = True
else:
print('WARNING - Not using GPU.')
pin_memory = False
test_loader = torch.utils.data.DataLoader(
def main():
train_z_transforms = transforms.Compose([
# RandomStretch(),
# CenterCrop((config.exemplar_size, config.exemplar_size)),
ToTensor()
])
train_x_transforms = transforms.Compose([
# RandomStretch(),
# RandomCrop((config.instance_size, config.instance_size),
# config.max_translate),
# ColorAug(config.color_ratio),
ToTensor()
])
val_z_transforms = transforms.Compose([
# CenterCrop((config.exemplar_size, config.exemplar_size)),
ToTensor()
])
val_x_transforms = transforms.Compose([
ToTensor()
])
score_size = int((config.instance_size - config.exemplar_size) / config.total_stride + 1)
anchors = generate_anchors(config.total_stride, config.anchor_base_size, config.anchor_scales,
config.anchor_ratios,
score_size)
# create dataset
train_dataset = GOT_10KDataset(train_data_dir, train_z_transforms, train_x_transforms, anchors)
valid_dataset = GOT_10KDataset(val_data_dir, val_z_transforms, val_x_transforms, anchors)
trainloader = DataLoader(train_dataset, batch_size=config.train_batch_size,
shuffle=True, pin_memory=True, num_workers=config.train_num_workers, drop_last=True)
validloader = DataLoader(valid_dataset, batch_size=config.valid_batch_size,
shuffle=False, pin_memory=True, num_workers=config.valid_num_workers, drop_last=True)
# create summary writer
if not os.path.exists(config.log_dir):
os.mkdir(config.log_dir)
summary_writer = SummaryWriter(config.log_dir)
# start training
with torch.cuda.device(config.gpu_id):
model = SiamRPN()
model.load_pretrain(pretrain_model_dir)
model.freeze_layers()
model = model.cuda()
optimizer = torch.optim.SGD(model.parameters(), lr=config.lr,
momentum=config.momentum, weight_decay=config.weight_decay)
# schdeuler = StepLR(optimizer, step_size=config.step_size, gamma=config.gamma)
scheduler = np.logspace(math.log10(config.lr), math.log10(config.end_lr), config.epoch)
for epoch in range(config.epoch):
train_loss = []
model.train()
curlr = scheduler[epoch]
for param_group in optimizer.param_groups:
param_group['lr'] = curlr
for i, data in enumerate(tqdm(trainloader)):
z, x, reg_label, cls_label = data
z, x = Variable(z.cuda()), Variable(x.cuda())
reg_label, cls_label = Variable(reg_label.cuda()), Variable(cls_label.cuda())
pred_cls, pred_reg = model(z, x)
optimizer.zero_grad()
# permute
pred_cls = pred_cls.reshape(-1, 1, config.anchor_num * score_size * score_size).permute(0,2,1)
pred_reg = pred_reg.reshape(-1, 4, config.anchor_num * score_size * score_size).permute(0,2,1)
cls_loss = rpn_cross_entropy_balance(pred_cls, cls_label, config.num_pos, config.num_neg)
reg_loss = rpn_smoothL1(pred_reg, reg_label, cls_label, config.num_pos)
loss = cls_loss + config.lamb * reg_loss
loss.backward()
torch.nn.utils.clip_grad_norm_(model.parameters(), config.clip)
optimizer.step()
step = epoch * len(trainloader) + i
summary_writer.add_scalar('train/loss', loss.data, step)
train_loss.append(loss.data.cpu().numpy())
train_loss = np.mean(train_loss)
valid_loss = []
model.eval()
for i, data in enumerate(tqdm(validloader)):
z, x, reg_label, cls_label = data
z, x = Variable(z.cuda()), Variable(x.cuda())
reg_label, cls_label = Variable(reg_label.cuda()), Variable(cls_label.cuda())
pred_cls, pred_reg = model(z, x)
pred_cls = pred_cls.reshape(-1, 1, config.anchor_num * score_size * score_size).permute(0, 2, 1)
pred_reg = pred_reg.reshape(-1, 4, config.anchor_num * score_size * score_size).permute(0, 2, 1)
cls_loss = rpn_cross_entropy_balance(pred_cls, cls_label, config.num_pos, config.num_neg)
reg_loss = rpn_smoothL1(pred_reg, reg_label, cls_label, config.num_pos)
loss = cls_loss + config.lamb * reg_loss
valid_loss.append(loss.data.cpu().numpy())
valid_loss = np.mean(valid_loss)
print("EPOCH %d valid_loss: %.4f, train_loss: %.4f, learning_rate: %.4f" %
(epoch, valid_loss, train_loss, optimizer.param_groups[0]["lr"]))
summary_writer.add_scalar('valid/loss',
valid_loss, epoch + 1)
torch.save(model.cpu().state_dict(),
"./models/siamrpn_{}.pth".format(epoch + 1))
model.cuda()
import sys
import lab_distribution
from custom_transforms import RGB2LAB, ToTensor
from network import Net
from plot import *
from logger import Logger
from tensorboardX import SummaryWriter
from torch.autograd import Variable
#data_dir = '/home/perrito/kth/DD2424/project/images/stl10_binary/train_X.bin'
data_dir = '../data/stl10/data/stl10_binary/train_X.bin'
ab_bins_dict = lab_distribution.get_ab_bins_from_data(data_dir)
ab_bins, a_bins, b_bins = ab_bins_dict['ab_bins'], ab_bins_dict[
'a_bins'], ab_bins_dict['b_bins']
transform = transforms.Compose([RGB2LAB(ab_bins), ToTensor()])
trainset = torchvision.datasets.ImageFolder(root='tmp_red_bird',
transform=transform)
trainloader = torch.utils.data.DataLoader(trainset,
batch_size=1,
shuffle=False,
num_workers=2)
cnn = Net(ab_bins_dict)
cnn.get_rarity_weights(data_dir)
criterion = cnn.loss
optimizer = optim.Adam(cnn.parameters(), weight_decay=.001)
# optimizer = optim.SGD(cnn.parameters(), lr=1e-2, momentum=0.9)
logger = Logger('./log')
logger.add_graph(cnn, image_size=96)
def main():
'''parameter initialization'''
args = parser.parse_args()
exp_name_dir = util.experiment_name_dir(args.experiment_name)
'''model on gpu'''
model = TrackerSiamRPN()
'''setup train data loader'''
name = 'GOT-10k'
assert name in ['VID', 'GOT-10k', 'All']
if name == 'GOT-10k':
root_dir = args.train_path
seq_dataset = GOT10k(root_dir, subset='train')
elif name == 'VID':
root_dir = '../data/ILSVRC'
seq_dataset = ImageNetVID(root_dir, subset=('train'))
elif name == 'All':
root_dir_vid = '../data/ILSVRC'
seq_datasetVID = ImageNetVID(root_dir_vid, subset=('train'))
root_dir_got = args.train_path
seq_datasetGOT = GOT10k(root_dir_got, subset='train')
seq_dataset = util.data_split(seq_datasetVID, seq_datasetGOT)
print('seq_dataset', len(seq_dataset))
train_z_transforms = transforms.Compose([ToTensor()])
train_x_transforms = transforms.Compose([ToTensor()])
train_data = TrainDataLoader(seq_dataset, train_z_transforms,
train_x_transforms, name)
anchors = train_data.anchors
train_loader = DataLoader(dataset=train_data,
batch_size=config.train_batch_size,
shuffle=True,
num_workers=config.train_num_workers,
pin_memory=True)
'''setup val data loader'''
name = 'GOT-10k'
assert name in ['VID', 'GOT-10k', 'All']
if name == 'GOT-10k':
root_dir = args.train_path
seq_dataset_val = GOT10k(root_dir, subset='val')
elif name == 'VID':
root_dir = '../data/ILSVRC'
seq_dataset_val = ImageNetVID(root_dir, subset=('val'))
elif name == 'All':
root_dir_vid = '../data/ILSVRC'
seq_datasetVID = ImageNetVID(root_dir_vid, subset=('val'))
root_dir_got = args.train_path
seq_datasetGOT = GOT10k(root_dir_got, subset='val')
seq_dataset_val = util.data_split(seq_datasetVID, seq_datasetGOT)
print('seq_dataset_val', len(seq_dataset_val))
valid_z_transforms = transforms.Compose([ToTensor()])
valid_x_transforms = transforms.Compose([ToTensor()])
val_data = TrainDataLoader(seq_dataset_val, valid_z_transforms,
valid_x_transforms, name)
val_loader = DataLoader(dataset=val_data,
batch_size=config.valid_batch_size,
shuffle=False,
num_workers=config.valid_num_workers,
pin_memory=True)
'''load weights'''
if not args.checkpoint_path == None:
assert os.path.isfile(
args.checkpoint_path), '{} is not valid checkpoint_path'.format(
args.checkpoint_path)
checkpoint = torch.load(args.checkpoint_path, map_location='cpu')
if 'model' in checkpoint.keys():
model.net.load_state_dict(
torch.load(args.checkpoint_path, map_location='cpu')['model'])
else:
model.net.load_state_dict(
torch.load(args.checkpoint_path, map_location='cpu'))
torch.cuda.empty_cache()
print('You are loading the model.load_state_dict')
elif config.pretrained_model:
checkpoint = torch.load(config.pretrained_model)
# change name and load parameters
checkpoint = {
k.replace('features.features', 'featureExtract'): v
for k, v in checkpoint.items()
}
model_dict = model.net.state_dict()
model_dict.update(checkpoint)
model.net.load_state_dict(model_dict)
#torch.cuda.empty_cache()
print('You are loading the pretrained model')
'''train phase'''
train_closses, train_rlosses, train_tlosses = AverageMeter(), AverageMeter(
), AverageMeter()
val_closses, val_rlosses, val_tlosses = AverageMeter(), AverageMeter(
), AverageMeter()
train_val_plot = SavePlot(exp_name_dir, 'train_val_plot')
model.adjust_lr(args.epoch_i)
for epoch in range(args.epoch_i, config.epoches):
model.net.train()
if config.fix_former_3_layers:
util.freeze_layers(model.net)
print('Train epoch {}/{}'.format(epoch + 1, config.epoches))
train_loss = []
with tqdm(total=config.train_epoch_size) as progbar:
for i, dataset in enumerate(train_loader):
closs, rloss, loss = model.step(epoch,
dataset,
anchors,
i,
train=True)
closs_ = closs.cpu().item()
if np.isnan(closs_):
sys.exit(0)
train_closses.update(closs.cpu().item())
train_rlosses.update(rloss.cpu().item())
train_tlosses.update(loss.cpu().item())
progbar.set_postfix(closs='{:05.3f}'.format(train_closses.avg),
rloss='{:05.5f}'.format(train_rlosses.avg),
tloss='{:05.3f}'.format(train_tlosses.avg))
progbar.update()
train_loss.append(train_tlosses.avg)
if i >= config.train_epoch_size - 1:
'''save model'''
model.save(model, exp_name_dir, epoch)
break
train_loss = np.mean(train_loss)
'''val phase'''
val_loss = []
with tqdm(total=config.val_epoch_size) as progbar:
print('Val epoch {}/{}'.format(epoch + 1, config.epoches))
for i, dataset in enumerate(val_loader):
val_closs, val_rloss, val_tloss = model.step(epoch,
dataset,
anchors,
train=False)
closs_ = val_closs.cpu().item()
if np.isnan(closs_):
sys.exit(0)
val_closses.update(val_closs.cpu().item())
val_rlosses.update(val_rloss.cpu().item())
val_tlosses.update(val_tloss.cpu().item())
progbar.set_postfix(closs='{:05.3f}'.format(val_closses.avg),
rloss='{:05.5f}'.format(val_rlosses.avg),
tloss='{:05.3f}'.format(val_tlosses.avg))
progbar.update()
val_loss.append(val_tlosses.avg)
if i >= config.val_epoch_size - 1:
break
val_loss = np.mean(val_loss)
train_val_plot.update(train_loss, val_loss)
print('Train loss: {}, val loss: {}'.format(train_loss, val_loss))
def main():
train_z_transforms = transforms.Compose([
RandomStretch(),
CenterCrop((config.exemplar_size, config.exemplar_size)),
ToTensor()
])
train_x_transforms = transforms.Compose([
RandomStretch(),
RandomCrop((config.instance_size, config.instance_size),
config.max_translate),
ToTensor()
])
val_z_transforms = transforms.Compose([
CenterCrop((config.exemplar_size, config.exemplar_size)),
ToTensor()
])
val_x_transforms = transforms.Compose([
ToTensor()
])
# create dataset
train_dataset = GOT_10KDataset(train_data_dir, train_z_transforms, train_x_transforms)
valid_dataset = GOT_10KDataset(val_data_dir, val_z_transforms, val_x_transforms, training=False)
trainloader = DataLoader(train_dataset, batch_size=config.train_batch_size,
shuffle=True, pin_memory=True, num_workers=config.train_num_workers, drop_last=True)
validloader = DataLoader(valid_dataset, batch_size=config.valid_batch_size,
shuffle=False, pin_memory=True, num_workers=config.valid_num_workers, drop_last=True)
# create summary writer
if not os.path.exists(config.log_dir):
os.mkdir(config.log_dir)
summary_writer = SummaryWriter(config.log_dir)
# start training
with torch.cuda.device(config.gpu_id):
model = SiamFCNet()
model.init_weights()
# model.load_state_dict(torch.load('./models/siamfc_30.pth'))
model = model.cuda()
optimizer = torch.optim.SGD(model.parameters(), lr=config.lr,
momentum=config.momentum, weight_decay=config.weight_decay)
# optimizer = torch.optim.Adam(model.parameters(), lr=config.lr)
schdeuler = StepLR(optimizer, step_size=config.step_size,
gamma=config.gamma)
for epoch in range(config.epoch):
train_loss = []
model.train()
for i, data in enumerate(tqdm(trainloader)):
z, x = data
z, x = Variable(z.cuda()), Variable(x.cuda())
outputs = model(z, x)
optimizer.zero_grad()
loss = model.loss(outputs)
loss.backward()
optimizer.step()
step = epoch * len(trainloader) + i
summary_writer.add_scalar('train/loss', loss.data, step)
train_loss.append(loss.data.cpu().numpy())
train_loss = np.mean(train_loss)
valid_loss = []
model.eval()
for i, data in enumerate(tqdm(validloader)):
z, x = data
z, x = Variable(z.cuda()), Variable(x.cuda())
outputs = model(z, x)
loss = model.loss(outputs)
valid_loss.append(loss.data.cpu().numpy())
valid_loss = np.mean(valid_loss)
print("EPOCH %d valid_loss: %.4f, train_loss: %.4f, learning_rate: %.4f" %
(epoch, valid_loss, train_loss, optimizer.param_groups[0]["lr"]))
summary_writer.add_scalar('valid/loss',
valid_loss, epoch + 1)
torch.save(model.cpu().state_dict(),
"./models/siamfc_{}.pth".format(epoch + 1))
model.cuda()
schdeuler.step()
self._pick_img_pairs(index)
self.open()
self._tranform()
regression_target, conf_target = self._target()
self.count += 1
return self.ret['train_z_transforms_rgb'], \
self.ret['train_x_transforms_rgb'], \
self.ret['train_z_transforms_ir'], \
self.ret['train_x_transforms_ir'], \
regression_target, \
conf_target.astype(np.int64)
def __len__(self):
return config.train_epoch_size * 64
if __name__ == "__main__":
root_dir = '/home/krautsct/RGB-T234'
seq_dataset_rgb = GOT10k(root_dir, subset='train_i')
seq_dataset_i = GOT10k(root_dir, subset='train_i', visible=False)
train_z_transforms = transforms.Compose([ToTensor()])
train_x_transforms = transforms.Compose([ToTensor()])
train_data = TrainDataLoaderRGBT(seq_dataset_rgb, seq_dataset_i,
train_z_transforms, train_x_transforms)
res = train_data.__getitem__(180)
print(res)