DeepLabV3Hparams#

class composer.models.DeepLabV3Hparams(initializers=<factory>, num_classes=None, backbone_arch='resnet101', backbone_weights=None, use_plus=True, sync_bn=True, ignore_index=-1, cross_entropy_weight=1.0, dice_weight=0.0)[source]#

YAHP interface for: composer_deeplabv3().

Parameters

num_classes (int) – Number of classes in the segmentation task.
backbone_arch (str, optional) – The architecture to use for the backbone. Must be either ['resnet50', 'resnet101']. Default: 'resnet101'.
backbone_weights (str, optional) – If specified, the PyTorch pre-trained weights to load for the backbone. Currently, only [‘IMAGENET1K_V1’, ‘IMAGENET1K_V2’] are supported. Default: None.
use_plus (bool, optional) – If True, use DeepLabv3+ head instead of DeepLabv3. Default: True.
sync_bn (bool, optional) – If True, replace all BatchNorm layers with SyncBatchNorm layers. Default: True.
initializers (List[Initializer], optional) – Initializers for the model. [] for no initialization. Default: [].