def __init__(self, cfg): #Slowfast constructor
"""
The `__init__` method of any subclass should also contain these
arguments.
Args:
cfg (CfgNode): model building configs, details are in the
comments of the config file.
"""
super(SlowFast, self).__init__()
#Standard line at the beginning of every NN definition. super(Classname, self).__init__()
self.norm_module = get_norm(cfg)
#The normalization used in our case is usually batch normalization. Normalizing the values within a certain range ([0, 1]) can help
#Speed up the convergence. Batch normalization does this for values in the hidden layer as well as the input layer. It basically allows each layer in a
#deep network to learn independently. Watch : https://www.youtube.com/watch?v=nUUqwaxLnWs
self.enable_detection = cfg.DETECTION.ENABLE
self.num_pathways = 3
#The number of pathways in the architecture, in this case we have 2 pathways, slow and fast.
self._construct_network(cfg)
#Function call
init_helper.init_weights(self, cfg.MODEL.FC_INIT_STD,
cfg.RESNET.ZERO_INIT_FINAL_BN)
def __init__(self, cfg):
"""
The `__init__` method of any subclass should also contain these
arguments.
Args:
cfg (CfgNode): model building configs, details are in the
comments of the config file.
"""
super(X3D, self).__init__()
self.norm_module = get_norm(cfg)
self.enable_detection = cfg.DETECTION.ENABLE
self.num_pathways = 1
exp_stage = 2.0
self.dim_c1 = cfg.X3D.DIM_C1
self.dim_res2 = (round_width(self.dim_c1, exp_stage, divisor=8)
if cfg.X3D.SCALE_RES2 else self.dim_c1)
self.dim_res3 = round_width(self.dim_res2, exp_stage, divisor=8)
self.dim_res4 = round_width(self.dim_res3, exp_stage, divisor=8)
self.dim_res5 = round_width(self.dim_res4, exp_stage, divisor=8)
self.block_basis = [
# blocks, c, stride
[1, self.dim_res2, 2],
[2, self.dim_res3, 2],
[5, self.dim_res4, 2],
[3, self.dim_res5, 2],
]
self._construct_network(cfg)
init_helper.init_weights(self, cfg.MODEL.FC_INIT_STD,
cfg.RESNET.ZERO_INIT_FINAL_BN)
def init_weights(self, init_std, ll_generator=None):
# Init all weights
init_helper.init_weights(self, init_std, False)
if ll_generator is not None:
self.head.rt_projection.weight.data.normal_(mean=0.0,
std=init_std,
generator=ll_generator)
self.head.rt_projection.bias.data.zero_()
def __init__(self, cfg):
super(X3D, self).__init__()
self.norm_module = get_norm(cfg)
self.enable_detection = cfg.DETECTION.ENABLE
self.num_pathways = 1
self._construct_network(cfg)
init_helper.init_weights(
self, cfg.MODEL.FC_INIT_STD, cfg.RESNET.ZERO_INIT_FINAL_BN
)
def __init__(self, cfg):
"""
The `__init__` method of any subclass should also contain these
arguments.
Args:
cfg (CfgNode): model building configs, details are in the
comments of the config file.
"""
super(SlowFastModel, self).__init__()
self.num_pathways = 2
self._construct_network(cfg)
init_helper.init_weights(self, cfg.MODEL.FC_INIT_STD,
cfg.RESNET.ZERO_INIT_FINAL_BN)
def __init__(self, cfg):
"""
The `__init__` method of any subclass should also contain these
arguments.
Args:
cfg (CfgNode): model building configs, details are in the
comments of the config file.
"""
super(CNNCatLSTM, self).__init__()
# self.norm_module = get_norm(cfg)
# self.enable_detection = cfg.DETECTION.ENABLE
self.num_pathways = 1
self._construct_network(cfg)
init_helper.init_weights(self, cfg.MODEL.FC_INIT_STD,
cfg.RESNET.ZERO_INIT_FINAL_BN)
def __init__(self, cfg):
"""
The `__init__` method of any subclass should also contain these
arguments.
Args:
cfg (CfgNode): model building configs, details are in the
comments of the config file.
"""
super(Student_SlowFast_MobileNetV3, self).__init__()
self.norm_module = nn.BatchNorm2d
self.enable_detection = cfg.DETECTION.ENABLE
self.num_pathways = 2
self.cfg = cfg
self._construct_network(cfg)
init_helper.init_weights(self, cfg.MODEL.FC_INIT_STD,
cfg.RESNET.ZERO_INIT_FINAL_BN)
def __init__(self, cfg):
"""
The `__init__` method of any subclass should also contain these
arguments.
Args:
cfg (CfgNode): model building configs, details are in the
comments of the config file.
"""
super(SlowFast, self).__init__()
self.extract_mstcn_features = cfg.TEST.EXTRACT_MSTCN_FEATURES
self.extract_features = cfg.TEST.EXTRACT_FEATURES
self.enable_detection = cfg.DETECTION.ENABLE
self.num_pathways = 2
self._construct_network(cfg)
init_helper.init_weights(self, cfg.MODEL.FC_INIT_STD,
cfg.RESNET.ZERO_INIT_FINAL_BN)
def __init__(self,
cfg,
detection_header=head_helper.ResNetRoIHead,
recognition_header=head_helper.ResNetBasicHead):
"""
The `__init__` method of any subclass should also contain these
arguments.
Args:
cfg (CfgNode): model building configs, details are in the
comments of the config file.
"""
super(SlowFastModel, self).__init__()
self.enable_detection = cfg.DETECTION.ENABLE
self.num_pathways = 2
self.detection_header = detection_header
self.recognition_header = recognition_header
self._construct_network(cfg)
init_helper.init_weights(self, cfg.MODEL.FC_INIT_STD,
cfg.RESNET.ZERO_INIT_FINAL_BN)
def __init__(self, cfg):
"""
The `__init__` method of any subclass should also contain these
arguments.
Args:
cfg (CfgNode): model building configs, details are in the
comments of the config file.
"""
self.temp_crops = cfg.SWAV_nmb_frame_views
self.spat_crops_list = cfg.SWAV_nmb_crops
self.spat_crops_size = cfg.SWAV_size_crops
self.shuffle = cfg.SWAV_shuffle
super(SlowFastSWAV, self).__init__()
self.norm_module = get_norm(cfg)
self.enable_detection = cfg.DETECTION.ENABLE
self.num_pathways = 2
self._construct_network(cfg)
init_helper.init_weights(self, cfg.MODEL.FC_INIT_STD,
cfg.RESNET.ZERO_INIT_FINAL_BN)
def __init__(self, cfg):
super(SlowFastBbox, self).__init__()
self.cfg = cfg
self.num_pathways = 2
self.slow_fast = SlowFast(cfg)
self.slow_fast.head = nn.Identity()
pool_size = _POOL1[cfg.MODEL.ARCH]
slow_pool_size = [
[
cfg.DATA.NUM_FRAMES // cfg.SLOWFAST.ALPHA // pool_size[0][0],
cfg.DATA.CROP_SIZE // 32 // pool_size[0][1],
cfg.DATA.CROP_SIZE // 32 // pool_size[0][2],
],
[
cfg.DATA.NUM_FRAMES // cfg.SLOWFAST.ALPHA // pool_size[0][0],
1,
1,
],
]
fast_pool_size = [
[
cfg.DATA.NUM_FRAMES // pool_size[1][0],
cfg.DATA.CROP_SIZE // 32 // pool_size[1][1],
cfg.DATA.CROP_SIZE // 32 // pool_size[1][2],
],
[
cfg.DATA.NUM_FRAMES // pool_size[1][0],
1,
1,
],
]
model_pool_size = [slow_pool_size, fast_pool_size]
if cfg.EPICKITCHENS.ROI_BRANCH == 1:
roi_type = [[0], [1]]
elif cfg.EPICKITCHENS.ROI_BRANCH == 0:
roi_type = [[1], [0]]
elif cfg.EPICKITCHENS.ROI_BRANCH == 2:
roi_type = [[0, 1], [1]]
elif cfg.EPICKITCHENS.ROI_BRANCH == 3:
roi_type = [[0, 1], [0]]
elif cfg.EPICKITCHENS.ROI_BRANCH == 4:
roi_type = [[0], [0, 1]]
elif cfg.EPICKITCHENS.ROI_BRANCH == 5:
roi_type = [[0, 1], [0, 1]]
else:
roi_type = [[0], [0]]
dim_in = [cfg.RESNET.WIDTH_PER_GROUP * 32] * len(roi_type[0]) + [
cfg.RESNET.WIDTH_PER_GROUP * 32 // cfg.SLOWFAST.BETA_INV
] * len(roi_type[1])
self.bbox_head = head_helper.ResNetBboxClassifierHead(
dim_in=dim_in,
num_classes=cfg.MODEL.NUM_CLASSES,
pool_size=model_pool_size,
roi_type=roi_type,
resolution=[[cfg.DETECTION.ROI_XFORM_RESOLUTION] * 2] * 2,
scale_factor=[
cfg.DATA.CROP_SIZE // cfg.DETECTION.ROI_XFORM_RESOLUTION
] * 2,
act_func="softmax",
aligned=cfg.DETECTION.ALIGNED,
)
self.roi_type = roi_type
init_helper.init_weights(self.bbox_head, cfg.MODEL.FC_INIT_STD,
cfg.RESNET.ZERO_INIT_FINAL_BN)
