retinanet_resnet50_fpn_v2(*, weights: Optional[torchvision.models.detection.retinanet.RetinaNet_ResNet50_FPN_V2_Weights] = None, progress: bool = True, num_classes: Optional[int] = None, weights_backbone: Optional[torchvision.models.resnet.ResNet50_Weights] = None, trainable_backbone_layers: Optional[int] = None, **kwargs: Any) → torchvision.models.detection.retinanet.RetinaNet¶
Constructs an improved RetinaNet model with a ResNet-50-FPN backbone.
The detection module is in Beta stage, and backward compatibility is not guaranteed.
retinanet_resnet50_fpn()for more details.
RetinaNet_ResNet50_FPN_V2_Weights, optional) – The pretrained weights to use. See
RetinaNet_ResNet50_FPN_V2_Weightsbelow for more details, and possible values. By default, no pre-trained weights are used.
progress (bool) – If True, displays a progress bar of the download to stderr. Default is True.
num_classes (int, optional) – number of output classes of the model (including the background)
ResNet50_Weights, optional) – The pretrained weights for the backbone.
trainable_backbone_layers (int, optional) – number of trainable (not frozen) layers starting from final block. Valid values are between 0 and 5, with 5 meaning all backbone layers are trainable. If
Noneis passed (the default) this value is set to 3.
**kwargs – parameters passed to the
torchvision.models.detection.RetinaNetbase class. Please refer to the source code for more details about this class.
The model builder above accepts the following values as the
RetinaNet_ResNet50_FPN_V2_Weights.DEFAULTis equivalent to
RetinaNet_ResNet50_FPN_V2_Weights.COCO_V1. You can also use strings, e.g.
These weights were produced using an enhanced training recipe to boost the model accuracy. Also available as
box_map (on COCO-val2017)
__background__, person, bicycle, … (88 omitted)
The inference transforms are available at
RetinaNet_ResNet50_FPN_V2_Weights.COCO_V1.transformsand perform the following preprocessing operations: Accepts
(B, C, H, W)and single
(C, H, W)image
torch.Tensorobjects. The images are rescaled to