def atom_resnet18_DeT(iou_input_dim=(256,256), iou_inter_dim=(256,256), backbone_pretrained=True, merge_type='mean'):
# backbones
backbone_net = backbones.resnet18(pretrained=backbone_pretrained)
backbone_net_depth = backbones.resnet18(pretrained=backbone_pretrained)
# Bounding box regressor
iou_predictor = bbmodels.AtomIoUNet(pred_input_dim=iou_input_dim, pred_inter_dim=iou_inter_dim)
net = ATOMnet_DeT(feature_extractor=backbone_net, feature_extractor_depth=backbone_net_depth, bb_regressor=iou_predictor, bb_regressor_layer=['layer2', 'layer3'],
extractor_grad=False, merge_type=merge_type)
return net
def segm_resnet18(segm_input_dim=(256, 256),
segm_inter_dim=(256, 256),
backbone_pretrained=True,
topk_pos=3,
topk_neg=3,
mixer_channels=2):
# backbone
backbone_net = backbones.resnet18(pretrained=backbone_pretrained)
# segmentation dimensions
segm_input_dim = (64, 64, 128, 256)
segm_inter_dim = (4, 16, 32, 64)
segm_dim = (64, 64) # convolutions before cosine similarity
# segmentation
segm_predictor = segmmodels.SegmNet(segm_input_dim=segm_input_dim,
segm_inter_dim=segm_inter_dim,
segm_dim=segm_dim,
topk_pos=topk_pos,
topk_neg=topk_neg,
mixer_channels=mixer_channels)
net = SegmNet(feature_extractor=backbone_net,
segm_predictor=segm_predictor,
segm_layers=['conv1', 'layer1', 'layer2', 'layer3'],
extractor_grad=False)
return net
def SBDT_resnet18(input_dim=(128, 256),
locator_inter_dim=(128, 256),
iou_input_dim=(256, 256),
iou_inter_dim=(256, 256),
backbone_pretrained=True):
# backbone
backbone_net = backbones.resnet18(pretrained=backbone_pretrained)
# Bounding box regressor
iou_predictor = bbmodels.AtomIoUNet(input_dim=input_dim,
pred_input_dim=iou_input_dim,
pred_inter_dim=iou_inter_dim)
# locator
location_predictor = locmodels.OnlineRRNet18(
input_dim=input_dim, pred_input_dim=locator_inter_dim)
# SBDTNet
net = SBDTNet(feature_extractor=backbone_net,
feature_layer=['layer2', 'layer3'],
bb_regressor=iou_predictor,
location_predictor=location_predictor,
extractor_grad=False)
return net
def steepest_descent_learn_filter_resnet18_newiou(filter_size=1, optim_iter=3, optim_init_step=1.0, optim_init_reg=0.01, output_activation=None,
classification_layer='layer3', backbone_pretrained=False, clf_feat_blocks=1,
clf_feat_norm=True, init_filter_norm=False, final_conv=False,
out_feature_dim=256, init_gauss_sigma=1.0, num_dist_bins=5, bin_displacement=1.0, test_loss=None,
mask_init_factor=4.0, iou_input_dim=(256,256), iou_inter_dim=(256,256),
jitter_sigma_factor=None, train_backbone=True):
# backbone
backbone_net = backbones.resnet18(pretrained=backbone_pretrained)
norm_scale = math.sqrt(1.0 / (out_feature_dim * filter_size * filter_size))
# classifier
clf_feature_extractor = clf_features.residual_basic_block(num_blocks=clf_feat_blocks, l2norm=clf_feat_norm,
final_conv=final_conv, norm_scale=norm_scale,
out_dim=out_feature_dim)
initializer = clf_initializer.FilterInitializerLinear(filter_size=filter_size, filter_norm=init_filter_norm, feature_dim=out_feature_dim)
optimizer = clf_optimizer.SteepestDescentLearn(num_iter=optim_iter, filter_size=filter_size, init_step_length=optim_init_step,
init_filter_reg=optim_init_reg, feature_dim=out_feature_dim,
init_gauss_sigma=init_gauss_sigma, num_dist_bins=num_dist_bins,
bin_displacement=bin_displacement, test_loss=test_loss, mask_init_factor=mask_init_factor)
classifier = target_clf.LinearFilter(filter_size=filter_size, filter_initializer=initializer,
filter_optimizer=optimizer, feature_extractor=clf_feature_extractor,
output_activation=output_activation, jitter_sigma_factor=jitter_sigma_factor)
# Bounding box regressor
bb_regressor = bbmodels.AtomIoUNet(pred_input_dim=iou_input_dim, pred_inter_dim=iou_inter_dim)
net = OptimTracker(feature_extractor=backbone_net, classifier=classifier, bb_regressor=bb_regressor,
classification_layer=classification_layer, bb_regressor_layer=['layer2', 'layer3'], train_feature_extractor=train_backbone)
return net
def uinet_resnet18(backbone_pretrained=True):
# backbone
backbone_net = backbones.resnet18(pretrained=backbone_pretrained,output_layers=['conv1', 'layer1', 'layer2', 'layer3', 'layer4'])
# Bounding box regressor
mask_predictor = UNET.U_Net()
net = UInet(feature_extractor=backbone_net, unet=mask_predictor,
extractor_grad=False)
return net
def atom_resnet18(iou_input_dim=(256,256), iou_inter_dim=(256,256), backbone_pretrained=True):
# backbone
backbone_net = backbones.resnet18(pretrained=backbone_pretrained)
# Bounding box regressor
iou_predictor = bbmodels.AtomIoUNet(pred_input_dim=iou_input_dim, pred_inter_dim=iou_inter_dim)
net = ATOMnet(feature_extractor=backbone_net, bb_regressor=iou_predictor, bb_regressor_layer=['layer2', 'layer3'],
extractor_grad=False)
return net
def klcedimpnet18(filter_size=1, optim_iter=5, optim_init_step=1.0, optim_init_reg=0.01,
classification_layer='layer3', feat_stride=16, backbone_pretrained=True, clf_feat_blocks=1,
clf_feat_norm=True, init_filter_norm=False, final_conv=True,
out_feature_dim=256, gauss_sigma=1.0,
iou_input_dim=(256, 256), iou_inter_dim=(256, 256),
detach_length=float('Inf'), alpha_eps=0.0, train_feature_extractor=True,
init_uni_weight=None, optim_min_reg=1e-3, init_initializer='default', normalize_label=False,
label_shrink=0, softmax_reg=None, label_threshold=0, final_relu=False, init_pool_square=False,
frozen_backbone_layers=()):
if not train_feature_extractor:
frozen_backbone_layers = 'all'
# Backbone
backbone_net = backbones.resnet18(pretrained=backbone_pretrained, frozen_layers=frozen_backbone_layers)
# Feature normalization
norm_scale = math.sqrt(1.0 / (out_feature_dim * filter_size * filter_size))
# Classifier features
clf_feature_extractor = clf_features.residual_basic_block(num_blocks=clf_feat_blocks, l2norm=clf_feat_norm,
final_conv=final_conv, norm_scale=norm_scale,
out_dim=out_feature_dim, final_relu=final_relu)
# Initializer for the DiMP classifier
initializer = clf_initializer.FilterInitializerLinear(filter_size=filter_size, filter_norm=init_filter_norm,
feature_dim=out_feature_dim, init_weights=init_initializer,
pool_square=init_pool_square)
# Optimizer for the DiMP classifier
optimizer = clf_optimizer.PrDiMPSteepestDescentNewton(num_iter=optim_iter, feat_stride=feat_stride,
init_step_length=optim_init_step,
init_filter_reg=optim_init_reg, gauss_sigma=gauss_sigma,
detach_length=detach_length, alpha_eps=alpha_eps,
init_uni_weight=init_uni_weight,
min_filter_reg=optim_min_reg, normalize_label=normalize_label,
label_shrink=label_shrink, softmax_reg=softmax_reg,
label_threshold=label_threshold)
# The classifier module
classifier = target_clf.LinearFilter(filter_size=filter_size, filter_initializer=initializer,
filter_optimizer=optimizer, feature_extractor=clf_feature_extractor)
# Bounding box regressor
bb_regressor = bbmodels.AtomIoUNet(pred_input_dim=iou_input_dim, pred_inter_dim=iou_inter_dim)
# DiMP network
net = DiMPnet(feature_extractor=backbone_net, classifier=classifier, bb_regressor=bb_regressor,
classification_layer=classification_layer, bb_regressor_layer=['layer2', 'layer3'])
return net
def dimpnet18(filter_size=1, optim_iter=5, optim_init_step=1.0, optim_init_reg=0.01,
classification_layer='layer3', feat_stride=16, backbone_pretrained=True, clf_feat_blocks=1,
clf_feat_norm=True, init_filter_norm=False, final_conv=True,
out_feature_dim=256, init_gauss_sigma=1.0, num_dist_bins=5, bin_displacement=1.0,
mask_init_factor=4.0, iou_input_dim=(256, 256), iou_inter_dim=(256, 256),
score_act='relu', act_param=None, target_mask_act='sigmoid',
detach_length=float('Inf'), frozen_backbone_layers=()):
# Backbone
backbone_net = backbones.resnet18(pretrained=backbone_pretrained, frozen_layers=frozen_backbone_layers)
# Feature normalization
norm_scale = math.sqrt(1.0 / (out_feature_dim * filter_size * filter_size))
# Classifier features
clf_feature_extractor = clf_features.residual_basic_block(num_blocks=clf_feat_blocks, l2norm=clf_feat_norm,
final_conv=final_conv, norm_scale=norm_scale,
out_dim=out_feature_dim)
# Initializer for the DiMP classifier
initializer = clf_initializer.FilterInitializerLinear(filter_size=filter_size, filter_norm=init_filter_norm,
feature_dim=out_feature_dim)
# Optimizer for the DiMP classifier
optimizer = clf_optimizer.DiMPSteepestDescentGN(num_iter=optim_iter, feat_stride=feat_stride,
init_step_length=optim_init_step,
init_filter_reg=optim_init_reg, init_gauss_sigma=init_gauss_sigma,
num_dist_bins=num_dist_bins,
bin_displacement=bin_displacement,
mask_init_factor=mask_init_factor,
score_act=score_act, act_param=act_param, mask_act=target_mask_act,
detach_length=detach_length)
# The classifier module
classifier = target_clf.LinearFilter(filter_size=filter_size, filter_initializer=initializer,
filter_optimizer=optimizer, feature_extractor=clf_feature_extractor)
# Bounding box regressor
bb_regressor = bbmodels.AtomIoUNet(pred_input_dim=iou_input_dim, pred_inter_dim=iou_inter_dim)
# DiMP network
net = DiMPnet(feature_extractor=backbone_net, classifier=classifier, bb_regressor=bb_regressor,
classification_layer=classification_layer, bb_regressor_layer=['layer2', 'layer3'])
return net
def L2dimpnet18(filter_size=1, optim_iter=5, optim_init_step=1.0, optim_init_reg=0.01,
classification_layer='layer3', feat_stride=16, backbone_pretrained=True, clf_feat_blocks=1,
clf_feat_norm=True, init_filter_norm=False, final_conv=True,
out_feature_dim=256, iou_input_dim=(256, 256), iou_inter_dim=(256, 256),
detach_length=float('Inf'), hinge_threshold=-999, gauss_sigma=1.0, alpha_eps=0):
# Backbone
backbone_net = backbones.resnet18(pretrained=backbone_pretrained)
# Feature normalization
norm_scale = math.sqrt(1.0 / (out_feature_dim * filter_size * filter_size))
# Classifier features
clf_feature_extractor = clf_features.residual_basic_block(num_blocks=clf_feat_blocks, l2norm=clf_feat_norm,
final_conv=final_conv, norm_scale=norm_scale,
out_dim=out_feature_dim)
# Initializer for the DiMP classifier
initializer = clf_initializer.FilterInitializerLinear(filter_size=filter_size, filter_norm=init_filter_norm,
feature_dim=out_feature_dim)
# Optimizer for the DiMP classifier
optimizer = clf_optimizer.DiMPL2SteepestDescentGN(num_iter=optim_iter, feat_stride=feat_stride,
init_step_length=optim_init_step, hinge_threshold=hinge_threshold,
init_filter_reg=optim_init_reg, gauss_sigma=gauss_sigma,
detach_length=detach_length, alpha_eps=alpha_eps)
# The classifier module
classifier = target_clf.LinearFilter(filter_size=filter_size, filter_initializer=initializer,
filter_optimizer=optimizer, feature_extractor=clf_feature_extractor)
# Bounding box regressor
bb_regressor = bbmodels.AtomIoUNet(pred_input_dim=iou_input_dim, pred_inter_dim=iou_inter_dim)
# DiMP network
net = DiMPnet(feature_extractor=backbone_net, classifier=classifier, bb_regressor=bb_regressor,
classification_layer=classification_layer, bb_regressor_layer=['layer2', 'layer3'])
return net
def atom_resnet18(iou_input_dim=(256, 256),
iou_inter_dim=(256, 256),
backbone_pretrained=True,
cpu=False):
# backbone
backbone_net = backbones.resnet18(
output_layers=['conv1', 'layer1', 'layer2', 'layer3'],
pretrained=backbone_pretrained)
# Bounding box regressor
iou_predictor = bbmodels.AtomIoUNet(pred_input_dim=iou_input_dim,
pred_inter_dim=iou_inter_dim,
cpu=cpu)
# if training CPU version, then need to fine-tune regressor
regressor_grad = True if cpu else False
net = ATOMnet(feature_extractor=backbone_net,
bb_regressor=iou_predictor,
bb_regressor_layer=['layer2', 'layer3'],
extractor_grad=False,
regressor_grad=regressor_grad)
return net
