build_backbone

lightning_pose.models.backbones.torchvision.build_backbone(backbone_arch: Literal['resnet18', 'resnet34', 'resnet50', 'resnet101', 'resnet152', 'resnet50_contrastive', 'resnet50_animal_apose', 'resnet50_animal_ap10k', 'resnet50_human_jhmdb', 'resnet50_human_res_rle', 'resnet50_human_top_res', 'resnet50_human_hand', 'efficientnet_b0', 'efficientnet_b1', 'efficientnet_b2', 'vits_dino', 'vitb_dino', 'vitb_imagenet', 'vitb_sam'], pretrained: bool = True, model_type: str = 'heatmap', **kwargs) → Tuple

Load backbone weights for resnets, efficientnets, and other models from torchvision.

Parameters:

• backbone_arch – which backbone version/weights to use

• pretrained – True to load weights pretrained on imagenet

• model_type – “heatmap” or “regression”

Returns:

tuple

• backbone: pytorch model

• num_fc_input_features (int): number of input features to fully connected layer