def __init__(self, classes, num_layers=101, pretrained=False, class_agnostic=False):
self.model_path = 'data/pretrained_model/resnet101_caffe.pth'
self.dout_base_model = 1024
self.pretrained = pretrained
self.class_agnostic = class_agnostic
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self, classes, pretrained=False, class_agnostic=False):
self.model_path = 'data/pretrained_model/vgg16_caffe.pth'
self.dout_base_model = 512
self.pretrained = pretrained
self.class_agnostic = class_agnostic
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
classes,
num_layers=101,
pretrained=False,
class_agnostic=False):
self.model_path = 'data/pretrained_model/resnet' + str(
num_layers) + '_caffe.pth'
self.dout_base_model = 1024
self.pretrained = pretrained
self.class_agnostic = class_agnostic
self.num_layers = num_layers
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
classes,
arch,
pretrained=False,
class_agnostic=False,
imagenet_weight=None):
self.dout_base_model = dout_base_model[arch]
self.pretrained = pretrained
self.class_agnostic = class_agnostic
self.arch = arch
self.imagenet_weight = imagenet_weight
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
classes,
pretrained=False,
class_agnostic=False,
teaching=False):
self.model_path = 'data/pretrained_model/vgg16_caffe.pth'
self.dout_base_model = 512
self.pretrained = pretrained
self.class_agnostic = class_agnostic
self.teaching = teaching
pooling_size = 7
_fasterRCNN.__init__(self, classes, class_agnostic, pooling_size,
teaching)
def __init__(self,
classes,
num_layers=101,
pretrained=False,
class_agnostic=False,
K=-1):
self.model_path = 'data/pretrained_model/resnet101_caffe.pth'
self.dout_base_model = 1024
self.pretrained = pretrained
self.class_agnostic = class_agnostic
self.K = K # if K > 1, transform FasterRCNN to take a stack of K images as input
_fasterRCNN.__init__(self, classes, class_agnostic, K=self.K)
def __init__(self,
classes,
num_layers=101,
pretrained=False,
freeze=False,
set_bn_fix=False,
embed_size=128):
self.dout_base_model = 1024
self.pretrained = pretrained
self.freeze, self.set_bn_fix = freeze, set_bn_fix
self.embed_size = embed_size
_fasterRCNN.__init__(self, classes, True)
def __init__(self,
classes,
num_layers=169,
pretrained=False,
class_agnostic=False,
imagenet_weight=None):
self.dout_base_model = dout_base_model[
num_layers] # depth of RCNN_base output. pattern1=1280 / pattern2=640 (in dense169)
self.pretrained = pretrained
self.class_agnostic = class_agnostic
self.num_layers = num_layers
self.imagenet_weight = imagenet_weight
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
classes,
num_layers=101,
pretrained=False,
class_agnostic=False,
pretrained_model=""):
self.model_path = pretrained_model
self.dout_base_model = 1024
self.pretrained = pretrained
self.class_agnostic = class_agnostic
self.num_layers = num_layers
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
classes,
num_layers=101,
pretrained=False,
class_agnostic=False,
rpn_type='normal'):
self.model_path = 'data/pretrained_model/resnet101_caffe.pth'
self.dout_base_model = 1024
self.pretrained = pretrained
self.class_agnostic = class_agnostic
self.rpn_type = rpn_type
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
classes,
num_layers=101,
pretrained=False,
class_agnostic=False,
shrink=1,
mimic=False):
self.shrink = shrink
self.pretrained = pretrained
self.class_agnostic = class_agnostic
self.dout_base_model = 768 // shrink
_fasterRCNN.__init__(self, classes, class_agnostic, shrink, mimic)
def __init__(self,
classes,
num_layers=101,
pretrained=False,
class_agnostic=False,
imagenet_weight=None):
self.dout_base_model = dout_base_model[num_layers]
self.pretrained = pretrained
self.class_agnostic = class_agnostic
self.num_layers = num_layers
self.imagenet_weight = imagenet_weight
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
action_classes,
obj_classes,
num_layers=101,
pretrained=False,
class_agnostic=False):
self.model_path = 'data/pretrained_model/resnet101_caffe.pth'
# self.model_path = 'data/pretrained_model/faster_rcnn_1_7_10021.pth'
self.dout_base_model = 1024
self.pretrained = pretrained
self.class_agnostic = class_agnostic
_fasterRCNN.__init__(self, action_classes, obj_classes, class_agnostic)
def __init__(self,
classes,
num_layers=101,
pretrained=False,
class_agnostic=False,
layer=101):
self.layer = layer
self.model_path = 'data/pretrained_model/resnet{}_caffe.pth'.format(
self.layer)
self.dout_base_model = 256 if self.layer in (18, 34) else 1024
self.pretrained = pretrained
self.class_agnostic = class_agnostic
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
classes,
num_layers=101,
pretrained=False,
class_agnostic=False):
self.model_path = os.path.join(
'data/pretrained_model/',
'resnet{:d}-caffe.pth'.format(num_layers))
self.dout_base_model = 1024
self.pretrained = pretrained
self.class_agnostic = class_agnostic
self.num_layers = num_layers
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
classes,
pretrained=False,
class_agnostic=False,
is_apn=False):
self.model_path = 'data/pretrained_model/vgg16_caffe.pth'
self.dout_base_model = 512
#self.mv_stride = 16
self.pretrained = pretrained
self.class_agnostic = class_agnostic
#self.is_apn = is_apn
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
classes,
num_layers,
pretrained=False,
class_agnostic=False,
imagenet_weight=None):
# self.model_path = 'data/pretrained_model/resnet101_caffe.pth'
self.dout_base_model = dout_base_model[num_layers]
self.pretrained = pretrained
self.class_agnostic = class_agnostic
self.num_layers = num_layers
self.imagenet_weight = imagenet_weight
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
classes,
ver='10',
pretrained=False,
class_agnostic=False,
imagenet_weight=None):
# dim of output from RCNN_base block
self.dout_base_model = dout_base_model[ver]
self.ver = ver
self.pretrained = pretrained
self.class_agnostic = class_agnostic
self.imagenet_weight = imagenet_weight
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
classes,
pretrained=False,
class_agnostic=False,
is_deconv=False,
num_filters=32):
self.model_path = 'data/pretrained_model/vgg16_caffe.pth'
self.dout_base_model = 512
self.pretrained = pretrained
self.class_agnostic = class_agnostic
self.is_deconv = is_deconv
self.num_filters = num_filters
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
classes,
num_layers=101,
pretrained=False,
class_agnostic=False,
pair_prob=None,
attr_prob=None):
self.model_path = 'data/pretrained_model/resnet101_caffe.pth'
self.dout_base_model = 1024
self.pretrained = pretrained
self.class_agnostic = class_agnostic
self.pair_prob = pair_prob
self.attr_prob = attr_prob
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
classes,
pretrained=False,
class_agnostic=False,
modal="rgb",
model_path="data/pretrained_model/vgg16_caffe.pth"):
#model_path="data/pretrained_model/vgg16_caffe.pth"):
self.model_path = model_path
self.dout_base_model = 512
self.pretrained = pretrained
self.class_agnostic = class_agnostic
#单双通道模态
self.modal = modal
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
classes,
num_layers=16,
pretrained=False,
class_agnostic=False,
imagenet_weight=None):
# self.model_path = 'data/pretrained_model/vgg16_caffe.pth'
self.dout_base_model = dout_base_model[
num_layers] # dim of output from RCNN_base block
self.num_layers = num_layers
self.pretrained = pretrained
self.class_agnostic = class_agnostic
self.imagenet_weight = imagenet_weight
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
classes,
pretrained=False,
class_agnostic=False,
lighthead=True):
self.dout_base_model = 576 # Output channel at Stage4
self.dout_lh_base_model = 576
self.class_agnostic = class_agnostic
self.pretrained = pretrained
_fasterRCNN.__init__(self,
classes,
class_agnostic,
lighthead,
compact_mode=True)
def __init__(self,
classes,
num_layers=101,
pretrained=False,
class_agnostic=False):
self.num_layers = num_layers
self.pretrained = pretrained
self.class_agnostic = class_agnostic
self.classes = classes
if num_layers >= 50:
self.dout_base_model = 1024
else:
self.dout_base_model = 256
_fasterRCNN.__init__(self, self.classes, self.class_agnostic)
def __init__(self,
classes,
pretrained=False,
class_agnostic=False,
teaching=False):
self.dout_base_model = 256
self.model_path = 'data/pretrained_model/alexnet_torch.pth'
self.pretrained = pretrained
self.class_agnostic = class_agnostic
self.teaching = teaching
# todo parametrizzare
self.n_frozen_layers = 10
print("N_Frozen_layers: " + str(self.n_frozen_layers))
pooling_size = 6
_fasterRCNN.__init__(self, classes, class_agnostic, pooling_size,
teaching)
def __init__(self,
classes,
num_layers=101,
pretrained=False,
class_agnostic=False):
self.model_path = 'data/pretrained_model/resnet101_caffe.pth'
print("res101")
#self.dout_base_model = 1024
#lhy
self.dout_base_model = 1280
self.pretrained = pretrained
self.class_agnostic = class_agnostic
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
classes,
num_layers=101,
model_path=None,
pretrained=False,
class_agnostic=False):
if model_path is None:
self.model_path = 'data/pretrained_model/resnet101_caffe.pth'
else:
self.model_path = model_path
self.dout_base_model = 1024
self.pretrained = pretrained
self.class_agnostic = class_agnostic
self.num_layers = num_layers
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
classes,
net,
num_layers=101,
pretrained=False,
class_agnostic=False):
self.model_path = 'data/pretrained_model/resnet101_caffe.pth'
if net == 'res18_3d':
self.dout_base_model = 256
else:
self.dout_base_model = 1024
self.pretrained = pretrained
self.class_agnostic = class_agnostic
if net == 'res18_3d':
_fasterRCNN3d.__init__(self, classes, class_agnostic)
else:
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
classes,
num_layers=101,
pretrained=False,
class_agnostic=False,
pretrained_path=''):
self.model_path = 'data/pretrained_model/resnet101_caffe.pth'
if num_layers != 18:
self.dout_base_model = 1024
else:
self.dout_base_model = 256
self.pretrained = pretrained
self.class_agnostic = class_agnostic
self.num_layers = num_layers
self.pretrained_path = pretrained_path
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(self,
classes,
num_layers=50,
relation_module=False,
pretrained=False,
class_agnostic=False):
self.dout_base_model = 1024
self.pretrained = pretrained,
self.class_agnostic = class_agnostic
self.relation_module = relation_module
_fasterRCNN.__init__(self,
classes,
class_agnostic,
relation_module=relation_module)
self.fc_dim = 16
self.emb_ft_dim = 64
self.relation_output_dim = 2048
def __init__(self,
classes,
num_layers=101,
pretrained=False,
class_agnostic=False,
**kwargs):
self.model_path = cfg.RESNET_PATH
self.dout_base_model = 1024
self.pretrained = pretrained
self.layers = num_layers
self.class_agnostic = class_agnostic
if self.layers == 50:
self.model_path = cfg.RESNET50_PATH
if self.layers == 152:
self.model_path = cfg.RESNET152_PATH
_fasterRCNN.__init__(self, classes, class_agnostic)
def __init__(
self,
classes,
layer,
pretrained_path=None,
class_agnostic=False,
):
self.pretrained_path = pretrained_path
self.class_agnostic = class_agnostic
self.dout_base_model = 256
self.layer = layer
self.dout_lh_base_model = 245
_fasterRCNN.__init__(self, classes, class_agnostic, compact_mode=True)