def train(args):
assert torch.cuda.is_available(), 'CUDA is not available.'
torch.backends.cudnn.enabled = True
torch.backends.cudnn.benchmark = True
print('Arguments : -------------------------------')
for name, value in args._get_kwargs():
print('{:16} : {:}'.format(name, value))
# Data Augmentation
mean_fill = tuple( [int(x*255) for x in [0.485, 0.456, 0.406] ] )
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
# train_transform = [transforms.AugTransBbox(1, 0.5)]
train_transform = [transforms.PreCrop(args.pre_crop_expand)]
train_transform += [transforms.TrainScale2WH((1024, 1024))]
train_transform += [transforms.AugHorizontalFlip(args.flip_prob)]
train_transform += [transforms.ToTensor()]
train_transform = transforms.Compose( train_transform )
# Training datasets
train_data = GeneralDataset(args.num_pts, train_transform, args.train_lists)
train_loader = torch.utils.data.DataLoader(train_data, batch_size=args.batch_size, shuffle=False, num_workers=args.workers, pin_memory=True)
net = Model(args.num_pts)
# print(len(net.children()))
#for m in net.children():
# print(type(m))
criterion = wing_loss(args)
optimizer = torch.optim.SGD(net.parameters(), lr=args.LR, momentum=args.momentum,
weight_decay=args.decay, nesterov=args.nesterov)
scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer, milestones=args.schedule, gamma=args.gamma)
net = net.cuda()
net = torch.nn.DataParallel(net)
print('--------------', len(train_loader))
for epoch in range(3):
break
for i , (inputs, target, mask, cropped_size) in enumerate(train_loader):
target = target.squeeze(1)
inputs = inputs.cuda()
target = target.cuda()
mask = mask.cuda()
prediction = net(inputs)
loss = criterion(prediction, target, mask)
nums_img = inputs.size()[0]
for j in range(nums_img):
temp_img = inputs[j].permute(1,2,0)
temp_img = temp_img.mul(255).numpy()
temp_img = cv2.cvtColor(temp_img, cv2.COLOR_RGB2BGR)
pts = []
for d in range(args.num_pts):
pts.append((target[j][0][2*d].item(), target[j][0][2*d+1].item()))
bbox = [int(index[0].item()) for index in meta]
#print(pts)
draw_points(temp_img, pts, (0, 255, 255))
#draw_points(temp_img, [(bbox[0],bbox[1])], (0, 0, 255))
#draw_points(temp_img, [(bbox[2],bbox[3])], (0, 0, 255))
cv2.rectangle(temp_img,(bbox[0],bbox[1]),(bbox[2],bbox[3]),(0,255,0),4)
cv2.imwrite('{}-{}-{}.jpg'.format(epoch,i,j), temp_img)
# assert 1==0
#if i > 5:
# break
for a, v in enumerate(train_data.data_value):
image = cv2.imread(v['image_path'])
meta = v['meta']
# bbox = v['bbox']
bbox = v['meta'].get_box()
pts = []
for d in range(args.num_pts):
pts.append((meta.points[0, d], meta.points[1, d]))
draw_points(image, pts, (0, 255, 255))
cv2.rectangle(image,(int(bbox[0]), int(bbox[1])),(int(bbox[2]), int(bbox[3])),(0,255,0),4)
cv2.imwrite('ori_{}.jpg'.format(a), image)
def main(args):
assert torch.cuda.is_available(), 'CUDA is not available.'
torch.backends.cudnn.enabled = True
torch.backends.cudnn.benchmark = True
prepare_seed(args.rand_seed)
logstr = 'seed-{:}-time-{:}'.format(args.rand_seed, time_for_file())
logger = Logger(args.save_path, logstr)
logger.log('Main Function with logger : {:}'.format(logger))
logger.log('Arguments : -------------------------------')
for name, value in args._get_kwargs():
logger.log('{:16} : {:}'.format(name, value))
logger.log("Python version : {}".format(sys.version.replace('\n', ' ')))
logger.log("Pillow version : {}".format(PIL.__version__))
logger.log("PyTorch version : {}".format(torch.__version__))
logger.log("cuDNN version : {}".format(torch.backends.cudnn.version()))
# General Data Argumentation
mean_fill = tuple( [int(x*255) for x in [0.485, 0.456, 0.406] ] )
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
print('The flip is : {}, rotate is {}'.format(args.arg_flip, args.rotate_max))
train_transform = [transforms.PreCrop(args.pre_crop_expand)]
train_transform += [transforms.TrainScale2WH((args.crop_width, args.crop_height))]
train_transform += [transforms.AugScale(args.scale_prob, args.scale_min, args.scale_max)]
if args.arg_flip:
train_transform += [transforms.AugHorizontalFlip(args.flip_prob)]
if args.rotate_max:
train_transform += [transforms.AugRotate(args.rotate_max)]
train_transform += [transforms.AugCrop(args.crop_width, args.crop_height, args.crop_perturb_max, mean_fill)]
train_transform += [transforms.ToTensor(), normalize]
train_transform = transforms.Compose( train_transform )
eval_transform = transforms.Compose([transforms.PreCrop(args.pre_crop_expand), transforms.TrainScale2WH((args.crop_width, args.crop_height)), transforms.ToTensor(), normalize])
assert (args.scale_min+args.scale_max) / 2 == args.scale_eval, 'The scale is not ok : {},{} vs {}'.format(args.scale_min, args.scale_max, args.scale_eval)
# Model Configure Load
model_config = load_configure(args.model_config, logger)
args.sigma = args.sigma * args.scale_eval
logger.log('Real Sigma : {:}'.format(args.sigma))
# Training Dataset
if args.regression:
train_data = GeneralDatasetForRegression(train_transform, args.data_indicator)
else:
train_data = GeneralDataset(train_transform, args.sigma, model_config.downsample, args.heatmap_type, args.data_indicator)
train_data.load_list(args.train_lists, args.num_pts, True)
train_loader = torch.utils.data.DataLoader(train_data, batch_size=args.batch_size, shuffle=True, num_workers=args.workers, pin_memory=True)
# Evaluation Dataloader
eval_loaders = []
if args.eval_vlists is not None:
for eval_vlist in args.eval_vlists:
if args.regression:
eval_vdata = GeneralDatasetForRegression(eval_transform, args.data_indicator)
else:
eval_vdata = GeneralDataset(eval_transform, args.sigma, model_config.downsample, args.heatmap_type, args.data_indicator)
eval_vdata.load_list(eval_vlist, args.num_pts, True)
eval_vloader = torch.utils.data.DataLoader(eval_vdata, batch_size=args.batch_size, shuffle=False,
num_workers=args.workers, pin_memory=True)
eval_loaders.append((eval_vloader, True))
if args.eval_ilists is not None:
for eval_ilist in args.eval_ilists:
if args.regression:
eval_idata = GeneralDatasetForRegression(eval_transform, args.data_indicator)
else:
eval_idata = GeneralDataset(eval_transform, args.sigma, model_config.downsample, args.heatmap_type, args.data_indicator)
eval_idata.load_list(eval_ilist, args.num_pts, True)
eval_iloader = torch.utils.data.DataLoader(eval_idata, batch_size=args.batch_size, shuffle=False,
num_workers=args.workers, pin_memory=True)
eval_loaders.append((eval_iloader, False))
# Define network
logger.log('configure : {:}'.format(model_config))
if args.regression:
net = obtain_model(model_config, args.num_pts)
else:
net = obtain_model(model_config, args.num_pts + 1)
assert model_config.downsample == net.downsample, 'downsample is not correct : {} vs {}'.format(model_config.downsample, net.downsample)
logger.log("=> network :\n {}".format(net))
logger.log('Training-data : {:}'.format(train_data))
for i, eval_loader in enumerate(eval_loaders):
eval_loader, is_video = eval_loader
logger.log('The [{:2d}/{:2d}]-th testing-data [{:}] = {:}'.format(i, len(eval_loaders), 'video' if is_video else 'image', eval_loader.dataset))
logger.log('arguments : {:}'.format(args))
opt_config = load_configure(args.opt_config, logger)
if hasattr(net, 'specify_parameter'):
net_param_dict = net.specify_parameter(opt_config.LR, opt_config.Decay)
else:
net_param_dict = net.parameters()
optimizer, scheduler, criterion = obtain_optimizer(net_param_dict, opt_config, logger)
logger.log('criterion : {:}'.format(criterion))
net, criterion = net.cuda(), criterion.cuda()
net = torch.nn.DataParallel(net)
last_info = logger.last_info()
if last_info.exists():
logger.log("=> loading checkpoint of the last-info '{:}' start".format(last_info))
last_info = torch.load(last_info)
start_epoch = last_info['epoch'] + 1
checkpoint = torch.load(last_info['last_checkpoint'])
assert last_info['epoch'] == checkpoint['epoch'], 'Last-Info is not right {:} vs {:}'.format(last_info, checkpoint['epoch'])
net.load_state_dict(checkpoint['state_dict'])
optimizer.load_state_dict(checkpoint['optimizer'])
scheduler.load_state_dict(checkpoint['scheduler'])
logger.log("=> load-ok checkpoint '{:}' (epoch {:}) done" .format(logger.last_info(), checkpoint['epoch']))
else:
logger.log("=> do not find the last-info file : {:}".format(last_info))
start_epoch = 0
if args.eval_once:
logger.log("=> only evaluate the model once")
eval_results = eval_all(args, eval_loaders, net, criterion, 'eval-once', logger, opt_config)
logger.close() ; return
# Main Training and Evaluation Loop
start_time = time.time()
epoch_time = AverageMeter()
for epoch in range(start_epoch, opt_config.epochs):
scheduler.step()
need_time = convert_secs2time(epoch_time.avg * (opt_config.epochs-epoch), True)
epoch_str = 'epoch-{:03d}-{:03d}'.format(epoch, opt_config.epochs)
LRs = scheduler.get_lr()
logger.log('\n==>>{:s} [{:s}], [{:s}], LR : [{:.5f} ~ {:.5f}], Config : {:}'.format(time_string(), epoch_str, need_time, min(LRs), max(LRs), opt_config))
# train for one epoch
if args.regression:
train_loss, train_nme = basic_train_regression(args, train_loader, net, criterion, optimizer, epoch_str, logger, opt_config)
else:
train_loss, train_nme = basic_train(args, train_loader, net, criterion, optimizer, epoch_str, logger, opt_config)
# log the results
logger.log('==>>{:s} Train [{:}] Average Loss = {:.6f}, NME = {:.2f}'.format(time_string(), epoch_str, train_loss, train_nme*100))
# remember best prec@1 and save checkpoint
save_path = save_checkpoint({
'epoch': epoch,
'args' : deepcopy(args),
'arch' : model_config.arch,
'state_dict': net.state_dict(),
'scheduler' : scheduler.state_dict(),
'optimizer' : optimizer.state_dict(),
}, logger.path('model') / '{:}-{:}.pth'.format(model_config.arch, epoch_str), logger)
last_info = save_checkpoint({
'epoch': epoch,
'last_checkpoint': save_path,
}, logger.last_info(), logger)
if args.regression:
eval_results = basic_eval_all_regression(args, eval_loaders, net, criterion, epoch_str, logger, opt_config)
else:
eval_results = basic_eval_all(args, eval_loaders, net, criterion, epoch_str, logger, opt_config)
# measure elapsed time
epoch_time.update(time.time() - start_time)
start_time = time.time()
logger.close()
def train(args):
assert torch.cuda.is_available(), 'CUDA is not available.'
torch.backends.cudnn.enabled = True
torch.backends.cudnn.benchmark = True
tfboard_writer = SummaryWriter()
logname = '{}'.format(datetime.datetime.now().strftime('%Y-%m-%d-%H:%M'))
logger = Logger(args.save_path, logname)
logger.log('Arguments : -------------------------------')
for name, value in args._get_kwargs():
logger.log('{:16} : {:}'.format(name, value))
# Data Augmentation
mean_fill = tuple([int(x * 255) for x in [0.485, 0.456, 0.406]])
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
train_transform = [
transforms.AugTransBbox(args.transbbox_prob, args.transbbox_percent)
]
train_transform += [transforms.PreCrop(args.pre_crop_expand)]
train_transform += [
transforms.TrainScale2WH((args.crop_width, args.crop_height))
]
#train_transform += [transforms.AugHorizontalFlip(args.flip_prob)]
#train_transform += [transforms.AugScale(args.scale_prob, args.scale_min, args.scale_max)]
#train_transform += [transforms.AugCrop(args.crop_width, args.crop_height, args.crop_perturb_max, mean_fill)]
if args.rotate_max:
train_transform += [transforms.AugRotate(args.rotate_max)]
train_transform += [
transforms.AugGaussianBlur(args.gaussianblur_prob,
args.gaussianblur_kernel_size,
args.gaussianblur_sigma)
]
train_transform += [transforms.ToTensor(), normalize]
train_transform = transforms.Compose(train_transform)
eval_transform = transforms.Compose([
transforms.PreCrop(args.pre_crop_expand),
transforms.TrainScale2WH((args.crop_width, args.crop_height)),
transforms.ToTensor(), normalize
])
# Training datasets
train_data = GeneralDataset(args.num_pts, train_transform,
args.train_lists)
train_loader = torch.utils.data.DataLoader(train_data,
batch_size=args.batch_size,
shuffle=True,
num_workers=args.workers,
pin_memory=True)
# Evaluation Dataloader
eval_loaders = []
for eval_ilist in args.eval_lists:
eval_idata = GeneralDataset(args.num_pts, eval_transform, eval_ilist)
eval_iloader = torch.utils.data.DataLoader(eval_idata,
batch_size=args.batch_size,
shuffle=False,
num_workers=args.workers,
pin_memory=True)
eval_loaders.append(eval_iloader)
net = Model(args.num_pts)
logger.log("=> network :\n {}".format(net))
logger.log('arguments : {:}'.format(args))
optimizer = torch.optim.SGD(filter(lambda p: p.requires_grad,
net.parameters()),
lr=args.LR,
momentum=args.momentum,
weight_decay=args.decay,
nesterov=args.nesterov)
scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer,
milestones=args.schedule,
gamma=args.gamma)
criterion = wing_loss(args)
# criterion = torch.nn.MSELoss(reduce=True)
net = net.cuda()
criterion = criterion.cuda()
net = torch.nn.DataParallel(net)
last_info = logger.last_info()
if last_info.exists():
logger.log("=> loading checkpoint of the last-info '{:}' start".format(
last_info))
last_info = torch.load(last_info)
start_epoch = last_info['epoch'] + 1
checkpoint = torch.load(last_info['last_checkpoint'])
assert last_info['epoch'] == checkpoint[
'epoch'], 'Last-Info is not right {:} vs {:}'.format(
last_info, checkpoint['epoch'])
net.load_state_dict(checkpoint['state_dict'])
optimizer.load_state_dict(checkpoint['optimizer'])
scheduler.load_state_dict(checkpoint['scheduler'])
logger.log("=> load-ok checkpoint '{:}' (epoch {:}) done".format(
logger.last_info(), checkpoint['epoch']))
else:
logger.log("=> do not find the last-info file : {:}".format(last_info))
start_epoch = 0
for epoch in range(start_epoch, args.epochs):
scheduler.step()
net.train()
# train
img_prediction = []
img_target = []
train_losses = AverageMeter()
for i, (inputs, target) in enumerate(train_loader):
target = target.squeeze(1)
inputs = inputs.cuda()
target = target.cuda()
#print(inputs.size())
#ssert 1==0
prediction = net(inputs)
loss = criterion(prediction, target)
train_losses.update(loss.item(), inputs.size(0))
prediction = prediction.detach().to(torch.device('cpu')).numpy()
target = target.detach().to(torch.device('cpu')).numpy()
for idx in range(inputs.size()[0]):
img_prediction.append(prediction[idx, :])
img_target.append(target[idx, :])
optimizer.zero_grad()
loss.backward()
optimizer.step()
if i % args.print_freq == 0 or i + 1 == len(train_loader):
logger.log(
'[train Info]: [epoch-{}-{}][{:04d}/{:04d}][Loss:{:.2f}]'.
format(epoch, args.epochs, i, len(train_loader),
loss.item()))
train_nme = compute_nme(args.num_pts, img_prediction, img_target)
logger.log('epoch {:02d} completed!'.format(epoch))
logger.log(
'[train Info]: [epoch-{}-{}][Avg Loss:{:.6f}][NME:{:.2f}]'.format(
epoch, args.epochs, train_losses.avg, train_nme * 100))
tfboard_writer.add_scalar('Average Loss', train_losses.avg, epoch)
tfboard_writer.add_scalar('NME', train_nme * 100,
epoch) # traing data nme
# save checkpoint
filename = 'epoch-{}-{}.pth'.format(epoch, args.epochs)
save_path = logger.path('model') / filename
torch.save(
{
'epoch': epoch,
'args': deepcopy(args),
'state_dict': net.state_dict(),
'scheduler': scheduler.state_dict(),
'optimizer': optimizer.state_dict(),
},
logger.path('model') / filename)
logger.log('save checkpoint into {}'.format(filename))
last_info = torch.save({
'epoch': epoch,
'last_checkpoint': save_path
}, logger.last_info())
# eval
logger.log('Basic-Eval-All evaluates {} dataset'.format(
len(eval_loaders)))
for i, loader in enumerate(eval_loaders):
eval_losses = AverageMeter()
eval_prediction = []
eval_target = []
with torch.no_grad():
net.eval()
for i_batch, (inputs, target) in enumerate(loader):
target = target.squeeze(1)
inputs = inputs.cuda()
target = target.cuda()
prediction = net(inputs)
loss = criterion(prediction, target)
eval_losses.update(loss.item(), inputs.size(0))
prediction = prediction.detach().to(
torch.device('cpu')).numpy()
target = target.detach().to(torch.device('cpu')).numpy()
for idx in range(inputs.size()[0]):
eval_prediction.append(prediction[idx, :])
eval_target.append(target[idx, :])
if i_batch % args.print_freq == 0 or i + 1 == len(loader):
logger.log(
'[Eval Info]: [epoch-{}-{}][{:04d}/{:04d}][Loss:{:.2f}]'
.format(epoch, args.epochs, i, len(loader),
loss.item()))
eval_nme = compute_nme(args.num_pts, eval_prediction, eval_target)
logger.log(
'[Eval Info]: [evaluate the {}/{}-th dataset][epoch-{}-{}][Avg Loss:{:.6f}][NME:{:.2f}]'
.format(i, len(eval_loaders), epoch, args.epochs,
eval_losses.avg, eval_nme * 100))
tfboard_writer.add_scalar('eval_nme/{}'.format(i), eval_nme * 100,
epoch)
logger.close()
os.makedirs(experiment_path)
if args.run:
# Load params
params = pickle.load(open(model_path + '/params.p', 'rb'))
# Load model
state_dict = torch.load('{}/model/{}_state_dict_best.pth'.format(
model_path, params['model']),
map_location=lambda storage, loc: storage)
model = load_model(params['model'], params)
model.load_state_dict(state_dict)
# Load ground-truth states from test set
test_loader = DataLoader(GeneralDataset(params['dataset'],
train=False,
normalize_data=params['normalize'],
subsample=params['subsample']),
batch_size=args.n_samples,
shuffle=args.shuffle)
data, macro_intents = next(iter(test_loader))
data, macro_intents = data.transpose(0, 1), macro_intents.transpose(0, 1)
# Sample trajectories
samples, macro_samples = model.sample(data,
macro_intents,
burn_in=args.burn_in)
# Save samples
samples = samples.detach().numpy()
pickle.dump(samples,
open(experiment_path + '/samples.p', 'wb'),
model.load_state_dict(state_dict)
else:
printlog('{:03d} {} {}'.format(args.trial, args.model, args.dataset))
printlog(model.params_str)
printlog(
'start_lr {} | min_lr {} | subsample {} | batch_size {} | seed {}'.
format(args.start_lr, args.min_lr, args.subsample, args.batch_size,
args.seed))
printlog('n_params: {:,}'.format(params['total_params']))
printlog('best_loss:')
printlog('############################################################')
# Dataset loaders
kwargs = {'num_workers': 1, 'pin_memory': True} if args.cuda else {}
train_loader = DataLoader(GeneralDataset(args.dataset,
train=True,
normalize_data=args.normalize,
subsample=args.subsample),
batch_size=batch_size,
shuffle=True,
**kwargs)
test_loader = DataLoader(GeneralDataset(args.dataset,
train=False,
normalize_data=args.normalize,
subsample=args.subsample),
batch_size=batch_size,
shuffle=True,
**kwargs)
############################# TRAIN LOOP #############################
best_test_loss = 0
# Create save destination
save_path = 'datasets/{}/data/examples'.format(args.dataset)
if not os.path.exists(save_path):
os.makedirs(save_path)
# Set params
params = {
'dataset' : args.dataset,
'normalize' : True,
'n_samples' : args.n_samples,
'burn_in' : 0,
'genMacro' : True
}
# Load ground-truth states from test set
test_loader = DataLoader(
GeneralDataset(params['dataset'], train=False, normalize_data=params['normalize'], subsample=1),
batch_size=args.n_samples, shuffle=args.shuffle)
data, macro_intents = next(iter(test_loader))
data, macro_intents = data.detach().numpy(), macro_intents.detach().numpy()
# Get dataset plot function
dataset = import_module('datasets.{}'.format(params['dataset']))
plot_func = dataset.animate if args.animate else dataset.display
for k in range(args.n_samples):
print('Sample {:02d}'.format(k))
save_file = '{}/{:02d}'.format(save_path, k)
plot_func(data[k], macro_intents[k], params=params, save_file=save_file)