feature_extractor = models.resnet34(pretrained=True)
posenet = PoseNet(feature_extractor,
droprate=dropout,
pretrained=True,
filter_nans=(args.model == 'mapnet++'))
if args.model == 'posenet':
model = posenet
elif args.model.find('mapnet') >= 0:
model = MapNet(mapnet=posenet)
else:
raise NotImplementedError
# loss function
if args.model == 'posenet':
train_criterion = PoseNetCriterion(sax=sax,
saq=saq,
learn_beta=args.learn_beta)
val_criterion = PoseNetCriterion()
elif args.model.find('mapnet') >= 0:
kwargs = dict(sax=sax,
saq=saq,
srx=srx,
srq=srq,
learn_beta=args.learn_beta,
learn_gamma=args.learn_gamma)
if args.model.find('++') >= 0:
kwargs = dict(kwargs, gps_mode=(vo_lib == 'gps'))
train_criterion = MapNetOnlineCriterion(**kwargs)
val_criterion = MapNetOnlineCriterion()
else:
train_criterion = MapNetCriterion(**kwargs)
feature_extractor = models.resnet34(pretrained=True)
posenet = PoseNet(feature_extractor,
droprate=dropout,
pretrained=True,
filter_nans=False)
if args.model == 'posenet':
model = posenet
elif args.model.find('mapnet') >= 0:
model = MapNet(mapnet=posenet)
else:
raise NotImplementedError
# loss function
if args.model == 'posenet':
train_criterion = PoseNetCriterion(sax=sax, saq=saq, learn_beta=True)
val_criterion = PoseNetCriterion()
elif args.model.find('mapnet') >= 0:
kwargs = dict(sax=sax,
saq=saq,
srx=srx,
srq=srq,
learn_beta=True,
learn_gamma=True)
train_criterion = MapNetCriterion(**kwargs)
val_criterion = MapNetCriterion()
else:
raise NotImplementedError
# optimizer
param_list = [{'params': model.parameters()}]
feat_dim=feature_dim)
else:
raise NotImplementedError
# loss function
loss_fn = nn.L1Loss() #default regression loss
if loss_fn_config in ['l2', 'mse']:
loss_fn = nn.MSELoss()
print('Using MSE Loss')
elif loss_fn_config in ['smoothl1', 'huber']:
loss_fn = nn.SmoothL1Loss()
print('Using Smooth L1 / Huber Loss')
if args.model == 'posenet':
train_criterion = PoseNetCriterion(t_loss_fn=loss_fn,
q_loss_fn=loss_fn,
sax=sax,
saq=saq,
learn_beta=args.learn_beta)
val_criterion = PoseNetCriterion()
elif args.model == 'semanticOutput':
train_criterion = SemanticCriterion()
val_criterion = SemanticCriterion()
elif 'mapnet' in args.model or 'semantic' in args.model or 'multitask' in args.model:
kwargs = dict(t_loss_fn=loss_fn,
q_loss_fn=loss_fn,
sax=sax,
saq=saq,
srx=srx,
srq=srq,
learn_beta=args.learn_beta,
learn_gamma=args.learn_gamma)