vis4d.model.seg.fcn_resnet

FCN Resnet Implementation.

Classes

FCNResNet([base_model, num_classes, resize])

FCN with ResNet basemodel for semantic segmentation.

class FCNResNet(base_model='resnet50', num_classes=21, resize=(520, 520))

FCN with ResNet basemodel for semantic segmentation.

FCN with ResNet basemodel, following torchvision implementation.

<https://github.com/pytorch/vision/blob/main/torchvision/models/ segmentation/fcn.py>_.

model: FCNResNet(base_model=”resnet50”)

• dataset: Coco2017

• recipe: vis4d/model/segment/FCNResNet_coco_training.py

• metrics:

  • mIoU: 62.52

  • Acc: 90.50

forward(images)

Forward pass.

Parameters:

images (torch.Tensor) – Input images.

Returns:

Raw model predictions.

Return type:

FCNOut

forward_test(images)

Forward pass for testing.

Parameters:

images (torch.Tensor) – Input images.

Returns:

Raw model predictions.

Return type:

FCNOut

forward_train(images)

Forward pass for training.

Parameters:

images (torch.Tensor) – Input images.

Returns:

Raw model predictions.

Return type:

FCNOut